Skip to main content

Full text of "Solid State Design For The Radio Amateur 1986"

See other formats











Solid State Design 
for the Radio Amateur 


By Wes Hayward, W7ZOI 
and 

Doug DeMaw, W1FB 



American Radio Relay League, Inc. 
Newington, CT 061 1 1 


Copyright t 1986 by 

The American Radio Relay League, Inc. 

Copyright secured under the Pan-American 
Convention 

International Copyright secured 

This work is Publication No 31 of the Radio 
Amateur's Library, published by the League. 
All rights reserved. No part ol this work may 
be reproduced in any form except by written 
permission of the publisher. All rights of 
translation are reserved. 

Printed in USA 

Quedan reservados todos los derechos 

Library of Congress Catalog Card Number: 
77-730-94 

$ 12.00 

Second Printing 



This book was first released in 1977 
as a theoretical and practical guide for 
the radio amateur interested in using 
solid-state devices in RF design work. It 
gained a large, immediate following not 
only among amateurs, but among 
professional RF designers as well. 

In this second printing, the occasional 
errors and omissions which inevitably 
creep into a work of this magnitude have 
been corrected, making the publication 
even more valuable to its intended 
audience. 

It is our hope that this book will 
provide today’s readers with a thorough 
understanding of a technology which 
has left its indelible mark on radio- 
communication. 

David Sumner, K1ZZ 
Executive Vice President 



Acknowledgment 


This book not only reflects the recent work of the writers, but 
also the assistance of others. Without their help the book would 
not have been easy to prepare. It is impossible to list all of those 
who contributed, but I would like to mention a few and express 
my gratitude to them. Assistance in the construction of many of 
the projects was provided by Terry White (KL7IAK), Jeff Damm 
(WA7MLH), and Deane Kidd (W7TYR). I am grateful for dis- 
cussions with members of TERAC (Tektronix Employee's Radio 
Amateur Club, K7AUO) and for the photography done by Denton 
Bramwell. Special thanks goes to Mike Metcalf. W7UDM. He not 
only provided assistance and advice, but offered a number of his 
designs for our use. Discussions with my professional colleagues 
in the Communications Division at Tektronix have been helpful and 
enlightening. Additional thanks go to Linley Gumm (K7HFD), 

Fred Telewski (WA7TZY), and Larry Lockwood (W7JBY). Mention 
should be made of the liberal policy at Tektronix which allowed me 
to use its test equipment and computer facilities to generate data 
which would not have been available otherwise. 

Special recognition is given to my friend and co-author, Doug 
DeMaw, W1 FB. His candid views of my circuits and his tolerance 
of my forthright reviews of his work have, hopefully, led to designs 
which reflect sound engineering practice and ease of duplication. 

Finally, I would like to express my deep appreciation for the 
patience and assistance given by my wife, Shon, and our sons, Ron 
and Roger. Not only did Shon devote several hundred hours of 
typing time to the project, but she maintained an attitude of 
understanding and encouragement toward the book. The boys 
willingly gave up my time that could have been spent with them. 
They even breadboarded a few of the circuits described I 

Wes Hayward. W7ZOI 

Beaverton, Oregon 


No book of this kind is possible without the good will and assistance 
of the many people who work in the electronics industry as pro- 
fessional engineers and technicians. In our effort to make this publi- 
cation useful and informative to the reader it was necessary to con- 
sult with numerous key people in the semiconductor manufacturing 
field. I would like to express my gratitude to the personnel at RCA, 
Motorola and National Semiconductor Corp. who provided direct 
consultation for some of my circuits, data sheets, booklets and 
engineering samples of their various solid-state components. 

Without the generosity of Bill Amidon of Amidon Assoc, the 
circuits which contain ferrite and powdered-iron toroid cores would 
not have been so numerous. International Crystal Mfg. Co. and 
John Beanland (G3BVU) of Spectrum International, were responsible 
for many of the components used in these circuits. 

I wish to recognize the contributions of personal time and 
materials received from several members of the ARRL hq. staff, 
and finally I want to acknowledge the many hours without compen- 
sation that were invested by co-author W7ZOI during tape-letter 
and telephone exchanges of technical data. His intense motivation 
to m8ke this an outstanding contribution to the amateur's technical 
library led to many debates between the authors, and subsequently, 
a volume which will expand the technical knowledge of the reader. 

Doug DeMaw, W1FB (ex-WICER) 
Co-Author 




Contents 


1 Semiconductors and the Amateur 

Page 7 

2 Basics of Transmitter Design 

Page 17 

3 More Transmitter Topics 

Page 32 

4 Power Amplifiers and Matching 
Networks 

Page 52 

5 Receiver Design Basics 

Page 69 

6 Advanced Receiver Concepts 

Page 111 

7 Test Equipment and Accessories 

Page 143 

8 Modulation Methods 

Page 181 

9 Field Operation, Portable Gear and 
Integrated Stations 

Page 209 

Appendix 

Page 236 

Bibliography 

Page 251 

Index 

Page 254 




Chapter 1 


Semiconductors 
and the Amateur 


Irom the start, amateur radio has been a 
pastime wherein those involved have 
communicated with one another by 
means of short waves, and at the offset 
via long-wave paths. During recent years 
much of the equipment built by ama- 
teurs has been for use at hf, vhf and 
above. Homemade gear has been as- 
sembled for two primary reasons - 
economics and the need for equipment 
with specific features or qualities not 
found in commercially manufactured 
amateur equipment. A third and impor- 
tant stimulus has been the amateur’s 
quest for knowledge of how circuits 
operate. Individual creative needs lure 
still others into the field of design, 
where the pride of achievement comes 
from the act of doing. Generally 
speaking, communication is for these 
fellows a means to an end not an 
end in itself. This volume is aimed at 
those amateurs who are not disposed to 
sitting in front of store-bought equip- 
ment and simply communicating with 
others who are similarly inspired. 

Emphasis is placed here on methods 
which are currently popular in the 
amateur community among experi- 
menters and designers. It is beyond the 
scope and size of this book to offer a 
complete treatment of solid-state design 
principles for communications, but in 
the broader sense the reader is referred 
to many general texts which treat most 
of the subjects covered here in some- 
what greater depth. For the most part, 
the topics treated in this publication are 
those which the authors have been 
involved with for the past several years 
while working with semiconductors as 
amateurs. All of the construction proj- 
ects illustrated herein have been built, 
tested and subjected to normal and 
sometimes stringent on-the-air use. Cir- 


cuits which are shown schematically, 
but which do not relate directly to a 
given construction project, are proven 
ones, and will provide good perfor- 
mance. 

Our present world of solid-state de- 
vice technology has been a springboard 
for experimenting amateurs in their 
development of simple and complex 
circuits for communications. The 
vacuum tube moves gradually into the 
shadows as the semiconductor advances 
in character and capability. Industrial 
designers are using transistors and ICs in 
nearly all applications where they per- 
form as good as or better than tubes, 
and in small-signal work transistors fill 
that role handily. Furthermore, the 
overall efficiency of a solid-state piece 
of equipment versus that of a com- 
parable unit employing vacuum tubes is 
markedly greater. Reliability is still an- 
other part of the design rationale when 
using semiconductors. Last, but definite- 
ly not least, practical miniaturization 
when semiconductors are used far sur- 
passes that which can be achieved with 
tubes. Amateurs have long been aware 
of the foregoing contrasts in active 
devices, and have forged ahead with 
enthusiasm as they designed and built 
transmitting and receiving equipment 
for (heir own use. This volume is in- 
tended as a guidepost for those ama- 
teurs who have embraced the tech- 
nology of solid-state circuit design. It is 
hoped that this primer in circuit design 
and application will serve as the basis 
for greater achievement by the reader, 
and that it will inspire further study and 
experimentation for many. 

Simplicity Versus Complexity 

In general, the writers have at- 
tempted to emphasize methods which 


are, at least conceptually, straight- 
forward. Frills have been incorporated 
only where they might serve specific 
needs in operating the equipment. In 
most cases the nonessential circuits can 
be deleted without causing a degrada- 
tion in overall utility. Such features as 
side-tone monitors, break-in delay TR 
switching, and VOX are among those 
frills being discussed. 

There is a tendency among some 
amateur experimenters to oversimplify 
their designs. That approach can lead to 
a piece of gear which does not function 
as desired. The equipment might even 
be plagued with spurious output and 
distortion. Designs are provided in this 
book which are clean in operation, and 
are generally more efficient than some 
of the most simple circuit configura- 
tions; e.g., tlie one-transistor crystal- 
controlled transmitter. 

Historically, amateurs have viewed 
the complexity of a piece of gear as 
being commensurate with the number 
of active devices in the circuit. For 
example, the five-tube receiver of the 
middle 1950s was considered by some a 
“simple design.” Conversely, those 15- 
and 20-tube multiconversion “super- 
hets” were regarded as complex pieces 
of station apparatus. Such a point of 
view is no longer appropriate, for nowa- 
days, the number of active devices has 
little bearing on the cost or complexity 
of a particular design. Most modern 
transistors are relatively inexpensive, as 
is true of ICs and diodes. One can view 
die addition of one or a few more 
solid-state devices to a circuit with the 
same casual oudook diat is taken when 
adding a resistor or capacitor. Indeed, in 
many instances die addition of active 
circuitry may allow the builder to leave 
out a collection of passive components. 


Semiconductors and the Amateur 


7 



Fig. 1 - Current flow in a diode versus the 
applied voltage. 


thereby enhancing miniaturization, low- 
ering cost and contributing to improved 
performance. Thus, counting the 
number of transistors or ICs in a circuit 
is not a recommended way of judging 
the simplicity of a circuit. 

Another matter of concern to the 
builder is being able to make the circuit 
perform correctly after it is built. Quite 
often a circuit which contains only a 
small number of components will work 
just as well as, or better than, a similar 
circuit which uses many more parts, or 
even some sophisticated integrated cir- 
cuits. There is irony in the fact that 
some simpler circuits will require adjust- 
ment by means of sophisticated labora- 
tory equipment in order to effect 
proper operation, while the seemingly 
more complex version may function 
perfectly when power is first applied. 
Casual observation should not be relied 
upon in the determination of circuit 
complexity. 

The Design Approach 

There are a number of techniques 
which can be used by the amateur or 
professional designer when building a 
piece of equipment. For many amateurs 
the approach has been purely an empiri- 
cal one. That is, the circuit must per- 
form a specific function, so the amateur 
tackles the assignment on an experi- 
mental basis. He may peruse the avail- 
able literature (application notes, data 
sheets, magazine articles) until he spots 
a circuit similar to what he has in mind. 
The circuit will be duplicated, except 
for subtle changes in component values. 
Then, measurements may be performed 
to discover whether or not the circuit 
functions “as advertised.” On the other 
hand, the professional engineer, if he is 
worldly wise in his field, will follow a 
totally different path. From the data 
sheets he will choose a device which 
appears to be appropriate for a given 
application. He will then design a circuit 
around the component, say, a transistor. 
He will utilize advanced analytical 
methods, often based on the availability 
of a computer. In this manner he will 
fully understand and establish the cir- 
cuit performance prior to building it. 


After the circuit is built in physical 
form, there is seldom a significant dif- 
ference between the predicted and 
actual performance. 

The two procedures just discussed 
are clearly extreme examples. Moreover, 
in the real world of electronics the two 
will merge. The more skilled amateur 
will engage in considerable analysis of 
his design before starting construction. 
As a result, he will spend less time to 
obtain proper circuit operation once the 
last wire has been soldered in place. In 
reality, a professional designer is likely 
to spend a great deal more time experi- 
menting with his circuits than we may 
suspect, and in particular where rf cir- 
cuits are concerned. Because of the 
experimental aspects of such work, ama- 
teur radio often serves as an excellent 
background for professional design ef- 
forts. 

In this book the authors attempt to 
approach solid-state design work from 
the middle ground. There are a number 
of circuits which can be “lifted” di- 
rectly for use in amateur applications. 
Regardless, an attempt is made to pro- 
vide straightforward mathematical pro- 
cedures and circuit models, both of 
which should enable the amateur de- 
signer/experimenter to gain a better 
understanding of the work he is under- 
taking. It is hoped that the fallout from 
his design work will assure improved 
equipment performance. 

Basic Transistor Modeling 

It is not appropriate now to include 
a detailed discussion of the solid-state 
physics which are the basis of transistor 
operation. The reader is referred to the 
series by Stoffels which appeared in 
QST , and which is available as a re- 
print. 1 It will serve as an excellent 
introductory treatise on the topics that 
will be highlighted in this book. In this 
section we will discuss some simplified 
“models” that can be used in the 
analysis of many communications cir- 
cuits. 

The term “model” may sound un- 
familiar when used in a commentary 
about electronics, even though we are 
familiar with the expression in other 
ways. Certainly, as youngsters most of 
us have built scaled-down models of 
aircraft, ships or cars. We not only 
ended up with an attractive replica of 
the item we were modeling, we learned 
something about the original after 
which the model was patterned, and in 
particular about its structure. 

Models are often used in the analysis 
of electronic circuits for the purpose of 
describing various components in terms 
of simpler and more basic circuit com- 
ponents. The junction diode serves as an 
excellent illustration of this method. A 

' Reprint available from ARRL for $1. 


X IDEAL 
■r DIODE 

I 


V 


Fig. 2 — Current flow in the "ideal" diode. 


physicist would examine a diode with 
bias provided from a battery and would 
proceed with a fairly complicated anal- 
ysis in order to describe the diode 
operation. First, he would describe the 
electric fields resulting from the applied 
voltage. Then he would proceed to 
calculate the density of electrons and 
holes within the semiconductor materi- 
al, the rate at which they are created 
(from knowledge of the material tem- 
perature), how the charges move 
through the material, and the rate at 
which they combine with one another. 
Such calculations would give him a 
rudimentary knowledge of what is hap- 
pening inside the diode. 

For the physicist or device engineer 
the preceding calculations (and many 
more) are significant. Were the circuit 
designer to go through such an exercise 
in analysis each time he wished to use a 
diode, he would be seriously en- 
cumbered. His only concern is with the 
behavior of the device when viewed 
from its two external terminals. 

The current flowing in a diode is 
given by the well-known diode equa- 
tion 


/ = /, (e<? V/kT -1) (Eq. 1A) 


where I s is the diode saturation current 
in amperes, V is the bias voltage across 
the diode, q is the fundamental elec- 
tronic charge, k is Boltzman’s constant 
and T is the temperature in degrees 
Kelvin. For room temperature (about 
300 degrees K), the fraction kT +q has 
the value of 26 millivolts. A germanium 
diode might have saturation currents in 
the neighborhood of 10”® A while a 
silicon diode would be typified by 
values closer to 10“' 3 A. This equation 
is plotted for a typical silicon diode in 
Fig. 1 . 

This information can be used di- 
rectly by the designer, and often it is. 
However, in many situations much less 
refined information is sufficient for 
design purposes. 

Fig. 2 illustrates a simplified version 
of the curve shown in Fig. 1 . This shows 
how the diode has been replaced by an 
“ideal” diode, the behavior of which 


8 Chapter 1 





Fig. 3 - Current flow in a perfect diode with 
offset. 


can be described easily. When the diode 
is reverse biased, there is absolutely no 
flow of current. However, when the 
diode is forward biased (a more positive 
potential applied to the p- than to the n- 
material of the diode), the current 
which flows is determined totally by the 
circuit external to the diode. The so- 
called perfect diode is a model we can 
use to describe the conduct of real 
diodes in many circuits. The use of a 
model leads to simplified analysis. An- 
other diode model is shown in Fig. 3, 
where a battery has been connected in 
series with a perfect diode. With a 
forward bias of approximately 0.6 volt, 
current will begin to flow, still being 
limited by the external circuitry. Ger- 
manium diodes start to conduct at a 
somewhat lower applied voltage, in the 
region of 0.2 to 0.4 volt. 

If two silicon diodes are connected 
back-to-back as shown in Fig. 4, a 
system behavior would prevail which 
could be analyzed using the model 
given. This arrangement provides a 
three-terminal device which looks 
strangely familiar. It resembles an npn 
bipolar transistor! Indeed, if an npn 
transistor were examined by means of 
an ohmmeter - connecting only two 
transistor terminals to the meter at one 
time - it would appear to be nothing 
but a pair of back-to-back diodes. 

A transistor, conversely, has a prop- 
erty which makes it quite different 
from a pair of isolated diodes. The 
characterization can be seen when one 
of the diodes within it (base-emitter 



Fig. 4— The basic transistor Is formed by 
bacK-to-bacK diodes. 


junction) is forward biased while the 
other (base-collector junction) is reverse 
biased. Under these conditions current 
will flow in the collector terminal! This 
would not occur when using a pair of 
reverse-connected diodes. 

Current flow in the collector is not 
highly dependent upon the voltage 
supplied to the collector. It is, however, 
quite dependent upon the current 
flowing in the base-emitter diode. This 
parameter is a relatively linear one - the 
collector current is directly proportional 
to the base current. The ratio of I c /J b is 
the beta of the transistor. 

Using the Information 

By using the foregoing information, 
we can construct a simple transistor 
model (Fig. 5). A new element has been 
introduced - the current generator. It is 
shown in a circle with an arrow which 
indicates the direction of current flow. 
The battery we used with our simplified 
silicon-diode model has been included in 
the base leg of the transistor model, for 
it is significant when describing transis- 
tor operation. A battery has been 
omitted in the collector circuit because 
the collector -base diode is reverse biased 
in the typical application. Amplification 
is implicit in this model, as the current 
generator in the collector represents not 
a constant current, but a dependent 
current where the pertinent inde- 
pendent variable is the base current. 

The model illustrated in Fig. 5 is not 
complete for many situations. If we 
backtrack momentarily to Fig. 1 , where 
a real diode is depicted, it can be seen 
that the current does not increase in- 
finitely as forward bias is applied. The 
current increase is sharp and pro- 
nounced with increasing voltage, but is 
finite in nature. This characteristic can 
be depicted in a transistor model by 
inserting a resistance in series with the 
base. The magnitude of this resistance 
can be given approximately by 


R b = (Eq. IB) 

‘e(dc) 


where the dc emitter current is in mA, 
R h is the base resistance in ohms, and 
(3 (beta) is the current gain introduced 
above. 

A matter of significance which is not 
covered in Fig. 5 is the frequency effect 
on transistor gain. It should be noted 
that at low frequencies beta is constant, 
with typical values ranging from 10 or 
20 to several hundred. However, as the 
operating frequency is increased in MHz 
the beta of the transistor tends to 
decrease. At an ac operating frequency 
called the f r of a transistor - some- 
times called the gain-bandwidth product 
- the beta (current gain) is unity, or 1. 



Fig. 5 — Initial transistor model. 


At operating frequencies below the ef- 
fective f T the current gain is often well 
approximated by /3 = / r +f op < where f T 
is the gain-bandwidth product and f 0 „ is 
the chosen frequency of operation. For 
example, a 2N3904 would have an 
effective beta of 10 at 30 MHz since its 
f T is 300 MHz. 

Fig. 6 shows a composite transistor 
model which is suitable for approximate 
analysis of circuits which employ bi- 
polar transistors at both low and high 
frequencies. This illustration is highly 
simplified. Models used by modern cir- 
cuit designers may contain a dozen or 
more elements instead of the few depic- 
ted in this example. It is not surprising 
that sophisticated methods lead to 
amazing accuracy in predicting actual 
circuit behavior. What is spectacular is 
the fact that for many routine kinds of 
circuits the simplified model of Fig. 6 
will provide surprisingly accurate results 
- often at very high frequencies. 

At low frequencies the beta of a 
2N3904 is 100 typically. Hence, if this 
transistor were biased for an emitter 
current of 10 mA, the base resistance, 
R b .would be 260 dims. 

Biasing of Bipolar Transistors 

The simplified model of a transistor 
presented in Fig. 6 can be used as a tool 
in the analysis of circuits such as ampli- 
fiers and switches. When a transistor is 


IDEAL 

0O-H— 

1 

fe,v 

- 

S 

< 

< 

1 



1 

P = Pdc at low/ op 


fr 

(3 = -jr— at high/ op 

Jop 



Fig. 6 — Transistor model used for circuit 

analysis at high and low frequencies. 


Semiconductors and the Amateur 


9 






used as an amplifier, it is usually biased 
with do voltages in such a way that the 
applied ac signals cause the existing 
(quiescent) dc currents and voltages 
associated with the transistor to be 
varied slightly. It is these variations that 
are usually of interest when an amplifier 
is built. 

In this section various methods for 
biasing bipolar transistors will be con- 
sidered. This will serve not only the 
purpose of reviewing these concepts, 
but will illustrate how the simple model 
can be used as a means of circuit 
analysis. 

As an example, a simple audio ampli- 
fier will be studied. A likely transistor 
for this application is the 2N3565 which 
has an f T of about 60 MHz and a dc 
beta of 100. In the example, the ampli- 
fier will be biased for a dc collector 
current of 1 mA with the emitter 
grounded and the collector at +6 volts. 
Shown in Fig. 7 is a possible amplifier 
circuit, a simplified version of the sche- 
matic diagram showing only the dc part 
of the circuit, and finally, the dc 
portion of the circuit with the simple 
model substituted for the more con- 
ventional transistor symbol. 

First of all, since the collector cur- 
rent is to be 1 mA, and the voltage at 



Fig. 7 - Representations for the analysis of a 
transistorized amplifier. 


tire collector +6 volts, the value of R c is 
determined. In this case, it is given by 


R e 


1 2V - 6V 
.001 A 


= 6000 ohms 

(Eq. 2) 


Further, knowing that the collector 
current is 1 mA, the base current to 
yield this value must be I mA/beta = 10 
/aA. Knowing this value, the net re- 
sistance in series with the base can now 
be determined. The value of R b was 
given earlier as 


Rh 


260 

Te (mA) 


= 2600 ohms 


(Eq. 3) 


The net resistance in series with the base 
will be 


R ne , = 12V -0.6V B 11.4V 
I0” S A 10' 5 A 


= 1.14 megohms 


(Eq. 4) 


R 1 is me rely this val ue less R b , or 1.1 37 
megohms. In practice the builder would 
probably take a one-megohm resistor 
from the parts box for use at R1 with 
minimal problems being encountered, 
assuming that the transistor parameters 
used in the calculation are accurate. 

In the real world, the biasing scheme 
outlined in Fig. 7 will sometimes work, 
but presents a number of problems. The 
main deficiency of such a design is that 
the dc beta of a given transistor type 
can vary considerably. For the 2N3565 
used in the example, a beta of 100 
might be typical, but values as high as 
300 are frequently encountered. As- 
suming that the value of beta is 300 and 
that a one-megohm resistor was used at 
R1 , the base current would be 1 1 .4 /iA 
and the collector current would tend to 
be 1 1 .4 X 10~ 6 X 300 = 3.42 mA. This 
much current flowing in the 6000-ohm 
collector resistor would lead to a voltage 
drop across the resistor of 20.5 volts, 
which might suggest that the collector 
voltage would be negative. This is not 
possible (because of the ideal diode 
built into the collector of the transistor 
model). In reality, the voltage of the 
collector will drop to zero, or ground, 
and then go no farther. The collector 
current will now be determined purely 
by R c , and in this case will be 2 mA 
instead of the 1 mA desired originally. 
Clearly, with the collector at ground 
potential, with excess base current 
keeping it there, the transistor is not 
going to function well as an amplifier. 
This condition, where (he collector 


voltage is less than the base voltage, is 
called saturation. The originally ana- 
lyzed case with the collector voltage 
larger titan that of the base is called the 
active region. 

The problems outlined above, which 
resulted from a beta that was higher 
than expected, can be circumvented by 
the use of other circuit configurations 
or the addition of other components. 
Shown in Fig. 8 is a variation which is 
still less than optimum but will at least 
ensure that the transistor is biased in the 
active region. Here, the voltage source 
used to drive the base-bias resistor is the 
collector of the transistor rather titan 
the 12-volt supply, as originally used. 
This arrangement has the advantage that 
negative feedback is applied to the base. 
That is, if the beta were higher titan the 
desired 100, this would cause the cur- 
rent in the transistor to increase beyond 
the 1-mA design goal. However, as the 
collector current increases, a larger IR 
drop occurs across R c , resulting in 
decreased collector voltage. This, in 
turn, decreases the base current, causing 
the collector voltage to stabilize at some 
value larger than zero, but still less than 
the desired 6 volts. The transistor will 
always be biased in the active region 
with this scheme. 

The reader might find it instructive 
to assume that the transistor beta is 200 
and analyze the circuit of Fig. 8 by 
using the simple model. The result for 
this problem is that V c = 3.94 V, l c = 
1.34 mA and I b = 6.68 /zA. (Hint : The 
solution of two simultaneous equations 
is required.) 

Shown in Fig. 9 is a circuit which is 
more typical of the techniques used for 
biasing transistors in well-designed 
amplifiers. In this scheme, the base is 
connected to a voltage divider formed 
by the 10,000- and 5.000-ohm resistors. 
A capacitor has been added from the 
emitter to ground. A capacitor has a 
characteristic that prevents the voltage 
impressed across it from changing in- 
stantaneously. Hence, for ac signals 
applied to the amplifier, the emitter 


+ 12 V 



Fig. 8 - Bias arrangement to ensure that the 
transistor is in the active region. 


10 Chapter 1 




may be regarded as being at ground 
potential. However, the dc voltage 
certainly will not be at ground. 

In Fig. 9B the dc part of the circuit 
has been drawn, omitting the details 
associated with the ac part of the 
amplifier. Using classic circuit theory, it 
may be shown that the voltage divider 
consisting of Rl and R2 may be re- 
placed with a lower voltage V in series 
with a resistance R' where 



Fig. 9 — Typical bias arrangement for a well- 

designed amplifier. 


r/'= y v R2 (Eq. 

V Vcc X Rl + R2 14 ' 


and R' is the parallel equivalent of Rl 
and R2. This equivalent circuit is shown 
in Fig. 9C. 

Presented in D of Fig. 9 is a sche- 
matic diagram which results when a 
simplified model of the transistor is 
substituted in the amplifier circuit. Note 
here that the model used is even simpler 
than the one employed earlier, and that 
the resistance of the base-bias divider, 
R' , has been omitted. These changes will 
be justified in the following text. 

Noting the equivalent circuit of Fig. 
9D, it can be seen that the emitter 
voltage is 0.6 lower than that of the 
base, or in this case, 3.4 volts. The dc 
current (lowing in the emitter is hence, 
by Ohm’s Law, 3.4 V -f 2000 ohms = 
1 .7 mA. We see from the model that the 
emitter current is the sum of the base 
and collector currents. However, the 
collector current = beta times the base 
current, and beta is typically a fairly 
high value. Thus, the emitter current is 
approximately equal to the collector 
current. Using this approximation, the 
collector current is also 1.7 mA. It is 
significant to note that the value of beta 
was not even used in the calculation of 
the emitter and collector currents. 

If the beta of the transistor used in 
the circuit of Fig. 9 was 100, the base 
current would be 1.7 mA/100 = 17 /rA. 
This current flow through R ' , the equiv- 
alent resistance of the R1-R2 voltage 
divider, would case a voltage drop of 
only .02 volt, causing the base voltage 
not to be 4 volts, but 3.98 volts. This is 
close enough to 4 volts that the more 
detailed calculation is not necessary. 
Generally speaking, the current flowing 
through the R1-R2 voltage divider (0.8 
mA in the example) should be large in 
comparison with the expected base cur- 
rent. As long as this constraint is main- 
tained, the simplified analysis is justi- 
fied. 

Throughout the text many circuits 
are presented, using this bias method, 
many of them containing dc voltage 
measurements at various points. The 
reader who is unfamiliar with biasing 
calculations is encouraged to use these 
examples as problems to test his under- 
standing of the foregoing concepts. 

Typically, the amateur designer 
biases his amplifiers with the thought 
that only a single power supply will be 
available - usually +12 volts. This con- 
straint is the result of the ultimate 
desire for using the gear in mobile or 
portable applications where only one 
power source is available. However, in 
modern industrial circuits it is common 
to find a number of power supplies 
available in a given piece of equipment. 
For example, in the typical Tektronix 



7000-series oscilloscope, voltages of 
+50, +15, +5, -15 and 50 volts are 
available to the designer. The access to a 
large number of supplies greatly simpli- 
fies design problems, especially where 
critical dc biasing situations are con- 
cerned. Shown in Fig. 10 is the method 
for biasing the simple amplifier just 
considered, when two supplies are avail- 
able. Since the base is virtually at dc 
ground potential, the emitter voltage is 
-0.6 volt. The emitter and, hence, the 
collector current are given approximate- 
ly by 



(Eq. 6) 


The collector voltage is merely V c = V cc 

- R C I c. 

A special type of diode, which is 
used frequently as a reference element 
in a voltage-regulator circuit, is the 
Zener diode. This component is merely 
a diode which is operated with a reverse 
bias that is allowed to increase until the 
reverse-diode breakdown potential is 
reached. This voltage is usually quite 
stable with temperature, and is rela- 
tively independent of the current 
flowing through the diode. Shown in 
Fig. 1 1 is a simple model for a Zener 
diode. 

Presented in Fig. 12 is a method for 
biasing a transistor amplifier when using 
a Zener diode. In the example, an 8-volt 
Zener diode is used, yielding l c = 1 mA, 
and V c = 6.6 volts. The approximate 
design equations are given in the figure. 



Fig. 1 1 — Zener diode model. 

Semiconductors and the Amateur 1 1 





+Vcc 



V c = 0.6 + V z - I C R C 


Fig. 1 2 - Amplifier bias using the Zener 
diode. 


Shown in Figs. 13 and 14 are two 
additional methods for biasing small- 
signal amplifiers. One scheme uses an- 
other transistor, in this case a pnp 
silicon device such as the 2N3906, while 
the other technique uses an inexpensive 
741 type of operational amplifier. The 
appropriate design equations are pre- 
sented with the figures. 

The last three biasing schemes may 
at first sight appear to be absurd, overly 
complicated and expensive. However, 
they all have a significant advantage 
which may not be apparent to the 
beginner. The asset is that the bias is 
quite stable and well regulated even 
though the emitter of the amplifier is at 
ground potential. This can be of ex- 
treme significance when the transistor 
must be operated at ultra-high fre- 
quencies (e.g., 1296 MHz), or if the 
amplifier is to be used as a relatively 
high-power output Class A amplifier at 
rf. In both of these situations it can be 
difficult to obtain suitable-quality 
bypass capacitors for the emitter which 
would allow the simpler methods out- 
lined in Fig. 9 to be used. Furthermore, 
tire transistors used in these applications 
may cost ten to twenty dollars. In such 
a situation, it is worth the investment of 
an extra dime for a Zener diode, a pnp 
transistor or a quarter for a 741 oper- 
ational amplifier. As outlined in an 
earlier section, the true complexity of a 
circuit is difficult to judge by casual 
observation. 

The Small-Signal Model 

The simple models presented in the 
preceding sections have been general 
purpose in that they can be used not 
only for the analysis of the dc biasing 



Fig. 13 - Separate transistor acting as a bias Fig. 14 - An operational amplifier supplying 
source. the bias voltage. 


conditions, but for the behavior of the 
amplifier with applied signals. The 
ability to do analysis at high frequencies 
was implicit in the model because tran- 
sistor beta was allowed to decrease lin- 
early with frequency, reaching unity at 
the fr of the transistor. The models 
used by the design engineer are much 
more complicated, often containing up- 
ward of -two dozen components, in- 
cluding many capacitive elements. The 
general procedures are, nonetheless, the 
same, although the mathematics are 
sufficiently complicated to require 
computer-based analysis at times. 

Even though the models presented 
above are quite simple when compared 
with those used by industry, further 
simplification can be realized if only 
small ac signals are considered in the 
analysis. As an example, consider die 
simple audio amplifier presented first in 
Fig. 9 and repeated in Fig. 15, with the 


circuit redrawn to include the general 
model. If this circuit is investigated, 
with respect now to the application of 
small ac signals, considerable simplifica- 
tion can be realized. 

Capacitors Cl and C2 serve as dc 
blocking units. That is, the dc voltage 
may be different between the two ter- 
minals of the capacitor. However, a 
small ac signal presented to one end of 
the capacitor will appear unattenuated 
at the other side of the capacitor. 
Similarly, capacitors C3 and C4 are 
included merely to insure that the emit- 
ter of the transistor and the power- 
supply terminal are at ground as far as 
ac signals are concerned. 

If die interior of the transistor 
model is investigated, a further re- 
duction can be realized. The 0.6-volt 
battery in series with die base may be 
eliminated, since small changes in base 
potential will be transmitted through 



Fig. 15 - The transistorized amplifier redrawn to include the transistor model. 


12 Chapter 1 







Fig. 16 - Small-signal model of the audio 
amplifier. 


the battery. Similarly, the ideal diode in 
the base is no longer of practical value, 
for the dc bias in the transistor will 
always keep this diode turned on as long 
as the input signals are kept small with 
respect to the dc levels present. Shown 
in Fig. 16 is the small-signal equivalent 
of the amplifier circuit of Fig. 15. 
Clearly, this circuit will be much easier 
to analyze than would be the case if the 
more complete model were used and all 
external components were retained. 

Consider that an ac input voltage of 
1-mV rms is applied to the circuit of 
Fig. 16. The input current will be E in 4 
R h . If the transistor has a beta at the 
operating frequency of 100 and is 
biased for 2 mA of emitter current, the 
input resistance of the transistor, R b , 
will be 1300 ohms. Hence, the current 
flowing into the base will be .001 V t 
1300 ohms = 0.77 pA. The current 
flowing into the collector will be beta 
times this value, or 77 microamps. If a 
2000-ohm load resistor. /?/., is used, the 
voltage across the resistor will be -l c X 
R l = -(77 X 10’ 6 X 2 X 10 3 ) = 
-0.1 5 4 V. The voltage gain is 154. 

The minus sign in the output is of 
significance. This can be seen from a 
close examination of the model. A 
current flowing into the base of the 
transistor leads to a larger current 
flowing into the collector. This current 
will flow through the load resistor in the 
direction indicated by the arrow. With 
one end of R[_ grounded, the current 
flow in the indicated direction will 
mean that the collector end of /?/, is 
going to be negative. Since we are 
dealing with ac signals, this minus sign 
indicates merely that the output voltage 
will be 180 degrees out of phase with 
die input voltage. 

Power delivered to a resistive load, 
R, is given as P = V 2 + R, where die 
voltage is die rms value. Using this 
equation, die input power delivered to 
die base is (.00 1) 2 / 1300 = 7.69 X 
10' 10 watt. The output power is simi- 
larly (0.1 54) J /2000 = 1.19 X I0' 5 
watt. The ratio of these powers is the 


power gain, in this case 1 5,400. This can 
be expressed in dB with the expression 
G p (dB) = 10 log PoutIPiH , or in this 
case 41.9 dB. 

The use of small-signal models is 
quite universal in almost all areas of 
circuit design, and the science has been 
well developed by using advanced ma- 
trix methods. This discipline is often 
described under the name “two-port 
network theory.” Although the math- 
ematics are complicated enough that 
such methods are not appropriate for a 
book aimed at the radio amateur, they 
are still exceedingly powerful, and do 
not require the use of a computer 
except in some of the more specialized 
cases. Some of the basic two-port net- 
work concepts are presented in the 
appendix, and have been used for many 
of the more refined designs in this book. 

Even though the full utilization of 
modeling methods is probably beyond 
some amateurs, the limited models can 
still be of extreme utility. When a 
circuit is first encountered, the builder 
should study the circuit and evaluate 
the biasing conditions. After this is 
done, the equivalent small-signal circuit 
may be redrawn, either on a sheet of 
paper or mentally. Through this process 
surprisingly complex circuits may often 
be analyzed with ease. 

Biasing and Modeling 
Field-Effect Transistors 

Although the workhorse of modern 
communications technology is the bi- 
polar transistor discussed in the pre- 
ceding sections, a device of increasing 
popularity is the field-effect transistor 
(FET). There are several methods which 
are used to construct FETs, leading to 
various schematic symbols and design 
approaches. The popularity of the FET 
with radio amateurs is, in large part, due 
to their similarity of behavior to the 
more familiar vacuum tube. 

The basic dc characteristics of an 
n-channel junction FET are outlined in 
Fig. 17. Probably the two most signifi- 
cant dc parameters are and V p . The 
current, lj ss , is that which will flow in 
the FET if the gate and source are tied 
together and the drain is biased at a 
voltage higher than the magnitude of 
Vp. The parameter V p is called the 
pinch-off voltage and is the voltage 
applied to the gate with respect to the 
source, which will cause the drain cur- 
rent to go virtually to zero. 

Probably the easiest method for de- 
signing the biasing of a JFET (junction 
FET) into the active region is to use a 
graphical technique to determine the 
value of a suitable source resistor. The 
circuit is shown in Fig. 18, and a 
suitable graph is shown in Fig. 19. In 
the graph we have assumed that the 
values for l dss and V p are, respectively, 
10 mA and —6 volts. The curve of Fig. 



Fig. 1 7 - Basic dc characteristics of the junc- 
tion FET. 


17 is approximated in the graph with a 
straight line. If it is desired to bias the 
FET to a drain current of 5 mA, a load 
line is drawn from the origin to the 
5-mA point on the FET characteristic 
curve. The voltage at this point is -3. 
The slope of this line is 3 V t 5 mA, 
corresponding to a resistance of 600 
ohms. This is thus the value of resistor 
which would be chosen for the source 
bias. While this method is approximate, 
it should suffice for most amateur ap- 
plications. 

Shown in Fig. 20 is a simple small- 
signal model for a JFET. Like the 
models used for the bipolar transistor, 
the basis which leads to a description of 
amplification is a dependent-current 
generator. However, where the bipolar 
transistor had a current generator in the 
collector circuit which was dependent 



Fig. 18 — FET biasing schematic. 


Semiconductors and the Amateur 13 






Fig. 19 — FET behavior with biasing. 


upon the base current, the generator in 
the FET is dependent upon the voltage 
on the gate of the FET. Since the input 
resistance of a typical FET is extremely 
high, the input can be fairly well repre- 
sented with an open circuit. The con- 
stant relating drain current to gate- 
source voltage is the transconductance 
and has the units of mhos (= 1 + ohms). 
Typical values might be 4000 micro- 
mhos, or .004 mho for a popular FET 
like the MPF102 or the 2N4416. 

Shown in Fig. 2 1 is a typical audio 
amplifier which uses an FET with the 
constants of the foregoing examples. In 
this circuit a large resistor is used to 
connect the gate of the FET to ground, 
to ensure that the proper bias condi- 
tions are maintained. Using the analysis 
methods just outlined, the dc drain 
voltage would be found to be +7, the dc 
source voltage would be +3, and the 
voltage gain would be 4. (Note that the 
transconductance of a typical bipolar 
transistor is much higher than that of an 
FET.) Although the voltage gain of the 
FET is only 4, the power gain is 
virtually infinite. This is because a finite 
power output is delivered to the 1000- 
ohm drain resistor, but the input to the 
FET is essentially an open circuit, which 
will not accept power. 

Negative Feedback and the 
Integrated-Circuit Operational 
Amplifier 

Although the transistors and FETs 
outlined in the previous sections are 


n ORAIN 



6 SOURCE 


SMALL- SIGNAL FET MODEL 


Fig. 20 — Small-signal model of JFET. 

14 Chapter 1 


used for the predominant applications 
in communications equipment, in many 
areas integrated circuits have gained 
wide acceptance. Of the many ICs avail- 
able, undoubtedly the most generally 
useful type is the operational amplifier, 
or "op-amp,” with the most common 
example being the /aA741. In recent 
years these devices have become so 
common in industry and in amateur 
work that their prices have dropped to 
very low levels. With such a low cost 
(usually 50 cents or less in small quanti- 
ties), they can be used with the same 
casualness that one would exercise in 
adding a transistor or a capacitor to a 
circuit. 

While 74 1 op amps have been used 
widely in amateur circles, they have also 
been used improperly in many situa- 
tions. The misuses have resulted from a 
lack of understanding of the principles 
and consequences of feedback and an 
incomplete understanding of a proper 
equivalent circuit to use in circuit design 
and analysis. 

Shown in Fig. 22 is the circuit 
symbol for an integrated op amp of the 
741 type along with a suitable equiv- 
alent circuit or model. There are several 
differences here from the models used 
with transistors and FETs. First, the 
output is not a current source, but a 
voltage source. Second, the op amp is a 
differential amplifier. That is, the out- 
put voltage is directly proportional to 
the difference between the two input 
voltages. The constant of proportion- 
ality is the open-loop voltage gain,.4 0 . 
Finally, the equivalent circuit of Fig. 22 
is reasonably accurate for both dc con- 
ditions and for small-signal analysis. 

The two inputs are labeled with a+ 
or a—. The + input means that an 
increase in the voltage at this terminal 
causes an increase in the output. This + 
terminal is called the nor.invcrting in- 
put. The — input, or the inverting input 
terminal, exhibits the opposite behavior. 
That is, an increase in its potential leads 
to a decrease in the output potential. 
The impedances seen at the two input 
terminals are high, typically. They are 
not as high as experienced with FETs, 
but are high enough to make the model 
of Fig. 22 valid in most applications. 

The value of A 0 is typically high - 
10,000 to 100,000, or even more. How- 
ever, this is the gain at dc and very low 
ac frequencies. As the frequency in- 
creases, the value of A a starts dropping, 
decreasing by a factor of two for every 
doubling of the frequency. The 741 op 
amp has a gain of approximately 1000 
at 1 kHz, and the voltage gain drops to 
unity at frequencies of about 500 kHz. 

There are some limitations to the 
performance of an op amp, and they are 
fairly obvious. Mainly, the output volt- 
age cannot go higher than the positive 
supply voltage, V cc , nor can it go lower 



Fig. 21 — Audio amplifier using a JFET. 


than V ee . Actually, with 741 -type op 
amps, one is safe to assume that the 
output can approach each supply within 
about 2 volts. If two supplies of + and 
-15 volts were used, as is the usual case 
with industrial equipment, the output 
might be expected to swing from —13 
to +13 volts. If a single 12-volt supply 
was used, as is the typical situation in 
most amateur applications, the output 
could be expected to range from +2 to 
+10 volts or a little higher. 

In discussing op amps, it is generally 
easier to describe the behavior if two 
supplies are used. Hence, for the typical 
amateur application where a single sup- 
ply is to be used, a “synthetic ground” 
will be created with a resistive divider. 
All voltages in the rest of this discussion 
will be with respect to this level. The 
circuit is shown in Fig. 23. Note that 
this would be exactly the same as 
working with + and -6-volt supplies, 
derived from a floating 1 2-volt battery. 

The behavior of an op amp will be 
described in terms of a number of 
circuit situations. The experimentally 
inclined amateur might wish to bread- 
board some of these in order to obtain a 
better feel for the phenomenon. 

In the first experiment (Fig. 24) the 
noninverting input of the amplifier is 
“grounded” and a signal, E ln , is applied 
through a 10-k£2 resistor to the in- 
verting input. The output is described 
by the equation, noting now that V+ = 0, 
leaving V out = -A 0 V mirws . Assume for 
this experiment that A a is 1000. If E 
were set at a positive 1 mV, the output 



Fig. 22 — Operational amplifier model. 







Fig. 23 — Synthetic ground for an operational 
amplifier. 


will be -1 volt. Similarly, if E were set 
at a negative 1 mV, the output would be 
1-volt positive with respect to the 
synthetic ground. 

It is also instructive to examine the 
input resistance of this composite ampli- 
fier. The op amp itself has virtually an 
open circuit at its input. Hence, no 
current will flow in the 10-kS2 resistor, 
and the resistance seen at the driving 
source, £, is essentially infinite. This 
may seem like a redundant statement at 
this point, but later experiments will 
lead to different results. 

Consider now the modification of 
the first experiment where a feedback 
resistor is added. This is presented in 
Fig. 25, where E is now +1 volt. As the 
input voltage is increased toward this 
1-volt level, the voltage at the inverting 
input will also tend to increase. This 
input change will be reflected through 
the amplifier and amplified by a factor 
of A ot making the output try to go 
negative. However, as the output voltage 
decreases, a negative voltage from the 
output is applied through the feedback 
resistor to the input. Since this fed-back 
input signal opposes the original driving 
signal, it is not immediately clear just 
where either the V minus input or the 
output voltage will end up. 

This is one of those situations where 
the use of a little elementary mathe- 
matics cannot be avoided. The pro- 
cedure in setting up the equations is 
really quite straightforward and should 
not frighten any amateur who has taken 
high-school algebra. 

Although the value is not yet known 
numerically, the voltage at the inverting 
input is specified as V minus . The current 
flowing into the overall circuit is(£- 
V minus)! Ri- Since the op amp itself appears 
as an open circuit, no current flows into 



its terminal. However, there will be 
current flowing in the feedback resistor 
with a magnitude of ( V minus - V oul )/R f . 
These two currents must be equal since 
the total current entering a point in a 
circuit must be zero. This gives us the 
equation 

~ V minus ) _ Knlnus ~ V out 

Ri Rf 

(Eq. 7) 



Fig. 25 - Operational amplifier with feedback. 


but, V oul is known: V out =a o (V+- 
Vminus) V minus- 

This value for V out is now substi- 
tuted in the first equation and the 
equation is solved for V minus . The net 
result is 

Vminus = - 

Ri(A 0 + \) + R f (Eq. 8) 

Noting again that V out = -A a V minus , we 
can solve for the closed-loop voltage 
gain. 



= -1-994 (Eq. 9) 

For large values of A a , we see that the 
last equation reduces to 


Gv~R f *R,m ?0k = 2 (Eq. 10) 

1 UK 

It is also instructive to calculate the 
input resistance of the circuit of Fig. 25. 
The effective input resistance is just R in 
= E Jr iin - But, the input current, is 
just given by the expression = E - 
Vminus * Ri where V minus was arrived 
at in an earlier equation. Using this 
expression and noting the values used in 
the diagram of A a = 1000, £, = 10 kS2 
and Rf = 20 kJ2. we calculate that the 
effective input resistance is 10,019.98 
ohms. Of this, 10,000 ohms is attribu- 
ted to the input resistor, R t . The other 
20 ohms is the effective resistance seen 
at the inverting input of the operational 
amplifier. Generally, the input resis- 
tance of such a circuit at the inverting 
input is R in at V minus port w R f -r A a . 
It can also be shown that the output 
resistance of an amplifier is reduced 
when negative feedback is introduced. 
To do this, we would have to modify 
our model to include some finite output 
resistance in series with the voltage 
source now used. 

While the foregoing analysis may 
appear to the amateur, who is uncom- 
fortable with simple mathematics, to be 
nothing but a bunch of esoteric gib- 
berish, the results are really profound 
and should be treated as such! In the 
beginning of the problem, we took an 
amplifier which had a high, but perhaps 
ill-defined, gain with input and output 


resistances which might be quite un- 
known. However, by applying feedback 
we ended up with a total circuit whose 
gain was determined by the ratio of two 
resistors and an input resistance which 
was well defined. Since the open-loop 
gain of the amplifier was variable with 
frequency, but the final expression for 
gain (Eq. 10) does not contain the 
open-loop gain, the ultimate amplifier 
response is virtually independent of 
frequency. 

There is another way to view the 
previous amplifier, which is extremely 
useful in the casual design of circuits 
with feedback. Viewing Fig. 25, while 
disregarding the mathematics for awhile, 
we see that the input signal causes a 
current to flow in /?,• and some small 
voltage to appear at the inverting input. 
However, with negative feedback the 
output voltage moves around in such a 
way that the voltage difference between 
the two inputs is maintained essentially 
at zero. 

This general view may be used to 
easily analyze a noninverting feedback 
amplifier. Consider the circuit shown in 
Fig. 26, where feedback is used but the 
input signal is applied to the nonin- 
verting input. With the input signal 
initially equal to zero, the output volt- 
age will adjust itself until the voltage at 
Vminus > s also zero. This will occur for 
V ou t = 0- Now, assume that E in is 
increased to 1 volt. The output voltage 
will move in such a manner that the 
voltage at V minus is also +1 volt. But, 
this will occur when the output voltage 
is 3 volts. The only place current can 
come from to put the inverting input at 
1 volt is from the divider formed by R f 
and Ri being fed by V out . In general the 



Fig. 26 - Non-inverting amplifier with feed- 
back. 


Semiconductors and the Amateur 15 








Fig. 27 - Transistorized amplifier with feed- 
back. 


gain of a non-inverting amplifier is 
Ci/«1+— for large A a (Eq. 11) 

K t 

Although it will not be shown at this 
time, feedback of this kind has the 
effect of increasing the input resistance 
seen at the non-inverting input, while 
still decreasing the output resistance. 
Again, these effects cannot be demon- 
strated mathematically with the model 
used due to the initial simplifying as- 
sumptions which were used. 

Although the details will not be 


presented until later chapters, feedback 
may be applied to simple one-transistor 
amplifiers in order to realize the same 
advantages achieved with an operational 
amplifier. Shown in Fig. 27 is the 
small-signal equivalent of a circuit of 
this kind. With the proper choice of 
feedback resistors, this amplifier may be 
designed such that the input and output 
impedances are both very close to 50 
ohms and the gain is flat from under 1 
MHz to the low vhf region if a good 
transistor is used. Feedback is one of 
the most powerful tools available to the 
amateur or professional designer. 


16 


Chapter 1 



Chapter 2 


Basics of Transmitter Design 


llie basic element of any amateur 
radio station is the transmitter. In years 
past, the transmitter found in the usual 
"ham shack” was a large unit, often 
mounted in a floor-to-ceiling rack cab- 
inet. This “machine” was decorated 
with a large collection of knobs and 
meters, all serving a necessary function. 
Some of the more elegant units even 
had windows which were covered with 
glass or a wire mesh, which allowed the 
final amplifier tubes to be monitored 
visually. Too much color on the plates 
indicated that perhaps the tubes were 
being pushed a little too hard. 

Times have changed and the modern 
homemade transmitter is often a small 
unit, designed with a minimum number 
of panel-mounted controls. If the 
builder acquires a flair for miniaturiza- 
tion, the QRP transmitter can be very 
small indeed. 

In spite of the variations in size, and 
the fact that most of the modern 
equipment built by the radio amateur is 
solid state, there are many similarities. 
Shown in Fig. 1 are block diagrams for 
cw transmitters of varying degrees of 
complexity. These range from the 
simple crystal-controlled transmitters to 
a frequency-synthesizer-based unit. All 
of these examples could be realized with 
modern solid-state technology or the 
vacuum-tube methods of the past. In 
this, as well as the following chapter, all 
of the systems outlined in the figures 
will be discussed. An attempt is made to 
expand those areas where minimum 
information has been published pre- 
viously. Many of the basics are reviewed 
also. 

Crystal Oscillators 

The workhorse of modern com- 
munications equipment is the crystal 


oscillator. In the simplest kind of trans- 
mitter, a crystal oscillator may serve as a 
complete circuit. More often, such oscil- 
lators are used to drive additional ampli- 
fiers to provide increased power output. 
In the more advanced amateur trans- 
mitters, crystal oscillators are used in 
conjunction with mixers and VFOs in a 
superheterodyne circuit design. Ulti- 
mately, the most advanced designs will 
use a crystal-controlled oscillator as the 
reference for a frequency synthesizer. 

The crystals used in communications 
technology are usually made from 
quartz, where the basis of operation is 
the piezoelectric effect. Materials which 
exhibit this effect have the character- 
istic that when subjected to an electric 
field, a mechanical stress occurs within 
(he crystalline material. The mechanical 
displacement resulting from this stress is 
often in a direction different from that 
ot the electric field. Depending upon 
the nature of the crystalline material 
and the physical size and mounting, a 
quartz crystal will exhibit mechanical 
resonances in much the same way that 
the strings of a musical instrument have 
mechanical resonances. The unusual 
characteristic of piezoelectric devices is 
that not only can an electric field cause 
a stress which will excite an internal 
mechanical resonance, but the presence 
of mechanical stress will generate an 
electric field. The net result with a 
quartz crystal is that we end up with a 
small device consisting of nothing more 
than a piece of quartz with two elec- 
trical connections which, electrically, 
behaves just like a tuned circuit. The 
equivalent circuit for a quartz crystal is 
shown in Fig. 2. 

The values associated with the equiv- 
alent L and C values are often much 
different than those we would ex- 


perience in circuits built with discrete 
components. For example, the series 
inductance, I. s , may approach one. 
henry, with a series capacitance of a few 
femtofarads (10" ls farad). The parallel 
capacitance, C p . is typically around 6 
pF. While not shown in the figure, there 
are also loss elements in a more com- 
plete equivalent circuit, which will give 
rise to a finite Q. The typical Q of a 
crystal which might be used in amateur 
transmitters would be around 50,000. 
In some special crystals, Qs of over 
1 ,000,000 are achieved. 

There are dozens of circuits which 
can be used to make oscillators with 
quartz crystals. We will present a few of 
them here. 

Shown in Fig. 3 is a circuit using a 
bipolar transistor. Here, a transistor is 
biased in the usual way, and is operated 
much like an LC tuned oscillator in the 
common-base mode. However, the usual 
base-bypass capacitor is replaced with a 
crystal which operates as a series-tuned 
circuit. With a 12-volt supply, this cir- 
cuit will deliver a typical power output 
of 20 mW or so. The signal on llie 
collector is approximately 10- to 15- 
volts pk-pk. 

In this oscillator stray and transistor 
internal capacitances provide feedback 
for oscillation. Proper feedback is main- 
tained by adjusting the external capaci- 
tor at the emitter of the transistor. This 
capacitor should be one which will 
exhibit some 200 ohms of reactance at 
die operating frequency (e.g., 100 pF at 
7 MHz). The tuned-collector circuit is 
resonant at the operadng frequency. 
This circuit may be hesitant about 
oscillating at the lower frequencies, 
especially at 160 and 80 meters. In 
diese cases, it is often possible to make 
an excellent oscillator by adding a ca- 

Basics of Transmitter Design 17 



Fig. 1 — Block diagrams of various cw-transmitter formats. 


pacitor between the base and the emit- multiples of the fundamental frequency, higher Q at its overtone frequencies 

ter. Typically, a capacitive reactance Furthermore, the high Q of a crystal (in than at the fundamental. 

(Xc) of 500 ohms is sufficient. comparison with that of a violin string) Shown in Fig. 4 is a simple crystal 

One useful characteristic of this cir- allows the overtone oscillation to occur oscillator using a junction field-effect 

cuit is that it will operate on the alone, without the presence of the transistor (JFET). This circuit will oper- 

overtone modes of a crystal. An over- fundamental. ate on crystal overtones as well as at the 

tone is merely an oscillation which uses An example of a third-overtone fundamental of the crystal, depending 
a harmonic resonance of the crystal, crystal oscillator is the circuit of Fig. 3 upon the tuning of the output circuit. 

That is, a violin string can be made to with all constants set for 21 MHz. The simplicity of this circuit makes it 

oscillate at frequencies higher than the However, the crystal is a 7-MHz funda- appealing, although the cost of a JFET 

one typically associated with the length mental unit. The output of tire overtone is usually higher than that of a good 

and tension in the string. It is the oscillator will be at 2 1 MHz. Absolutely bipolar transistor, 

existence of these harmonics, along with no output will be detected at 7 MHz! The JFET oscillator is converted 

the fundamental, which adds character When crystals are purchased, they easily to a simple variable-crystal oscil- 
to the sound, differentiating the violin will usually be fundamental-mode lator (VXO) by paralleling the crystal 

from a simple audio oscillator. In a devices up to a frequency of around 20 with a 100-pF variable capacitor. The 

similar manner, a crystal can be made to MHz. From 20 to 60 MHz, third- ability to “pull” the frequency of a 

oscillate on higher overtones. Because of overtone units are typical. Some Stir-, crystal is, generally, limited to funda- 

the mechanical boundary conditions 7th- and even 9th-overtone crystals are mental-mode oscillations in this circuit, 

imposed upon the crystal, overtone used in communications equipment. In Using a 14-MHz fundamental-mode 

oscillations will occur only at odd many cases a crystal will exhibit a crystal (International Crystal, type EX), 


18 Chapter 2 









Fig. 2 - Equivalent circuit for a quartz crystal. 


a frequency shift of 8.4 kHz was 
measured. On the other hand, using a 
7-MHz crystal, only 1.4 kHz of shift was 
measured. Although the ability to 
"VXO" a crystal is highly dependent 
upon individual crystal characteristics, 
the technique is still useful. For ex- 
ample, an oscillator like that shown in 
Fig. 4, operating at 18 MHz and fol- 
lowed by a suitable frequency -multiplier 
chain, could yield an excellent exciter 
for 2-meter cw. That approach could be 
used for the hf bands also, even though 
the tuning range would be limited. 

The bipolar-transistor oscillator of 
Fig. 3 can also be pulled by means of 
external components. This is most easily 
done by adding an inductor in series 
with die crystal. The inductance value 
will depend upon die individual crystal 
and the "pull" amount desired, but is 
typically a few microhenries (/jH) per 
kHz of shift when using a 7-MHz crystal. 
A simple means of utilizing this VXO 
capability is to mount a slide switch 
across the inductor. This will, in effect, 
give the builder the ability to shift his 
oscillator frequency down enough to 
dodge QRM, certainly a desired ob- 
jective with a crystal-controlled QRP 
transmitter as the example. Up to 15 
kHz of shift in a 7-MHz crystal- 
controlled oscillator has been measured 
with this circuit. 

Shown in Fig. 5 is a JFET VXO. In 
this circuit the system is optimized for 
maximum frequency shift with standard 
crystal types, while maintaining a fairly 
constant output voltage. This required 
die use of a dual-section variable 
capacitor for tuning, and careful com- 
ponent mounting was necessary to mini- 
mize stray capacitance. The inductor is 


a high-£? slug-tuned unit. Probably, the 
Q of the coil is not as critical as is the 
self-capacitance. A toroidal inductor on 
a relatively high permeability pow- 
dered-iron core (such as die Amidon 
Assoc. F series) might work well. Ex- 
perimentation is clearly required on die 
part of die builder. A frequency shift of 
12 kHz with a 6-MHz crystal, and a shift 
of 23 kHz with an 11-MHz unit was 
obtained, confirming that the maximum 
shift available is around 0.2 percent of 
the crystal frequency. A VXO of this 
kind would provide the basis for a 
number of interesting transmitters or 
transceivers. 

The VXO of Fig. 5 was bread- 
boarded and tested with a number of 
different crystals. An experimental 
change from the circuit shown was the 
use of a hot-carrier diode in place of the 
1N914 and smaller inductance values at 
L. The output is surprisingly constant 
over most of the tuning range of a given 
crystal, with variations less than 1 dB 
being typical. Using a 10-MHz crystal, a 
1 7-kl Iz shift was measured with a I6-/1II 
slug-tuned inductor. Several overtone 
crystals were operated on their funda- 
mental modes, and spectacular results 
were noted in some cases. For example, 
a 54-MHz third-overtone crystal was 
operated at 18 MHz with the 16-jrl I 
inductor. An excess of 150 kHz of shift 
was noted! The tuning was nonlinear, 
with most of the range being com- 
pressed near the low-C end of the variable 
capacitor spread. 

Two more oscillators using bipolar 
transistors are shown in Fig. 6. Neg- 
lecting slight differences in biasing, the 
circuits are essentially identical. They 
offer the advantage of requiring no 
tuned circuit for operation. Both are 
fundamental-mode oscillators. 

All of the circuits shown are aimed 
at reasonable stability, but have rela- 
tively low output power. It is possible 
to bias many of these circuits higher to 
obtain outputs of up to perhaps 1/4 
watt. However, thermal stability is often 
severely degraded, chirp is introduced if 
the oscillator is keyed, and the user 



OSCILLATOR 

MPFIOZ 



Fig. 4 — Crystal oscillator which employs a 
JFET. 


stands a chance of damaging the crystal 
from excessive rf current. It is not 
recommended that a single oscillator 
stage be used as a simple transmitter. 
The addition of an amplifier is so 
straightforward, and the system ef- 
ficiency is so much better, that the 
minimal simplicity is not of value. 

Most crystal oscillators which use 
bipolar transistors will operate fairly 
well with hundreds of different tran- 
sistor types. Generally, the only re- 
quirement other than the usual voltage- 
breakdown and maximum-current cri- 
terion is that the transistor have as high 
an f T as possible. This is met easily for 
oscillators in the hf region with tran- 
sistors likethe 2N3904, 2N4I24, 2N706, 
2N2222A, 2N3563 and others. For 
overtone oscillators operating well into 
the vhf region, one should select tran- 
sistors with an fj of I GHz or higher. 
The 2N5179 is excellent in such appli- 
cations. 

Designing Untuned Buffer Amplifiers 

While the output of a low-frequency 
crystal oscillator may be as high as 50 
milliwatts (mW) or more, the output 
from a VFO or mixer in a heterodyne 
exciter may be much less. An amplifier 
is needed to build up the power. Also, 
amplifiers help isolate an oscillator from 
the effects of changing load, such as 
might result from keying or modulation. 
These chores are usually handled by 
means of a Class A buffer/amplifier. In 
this section, the basics of untuned 
amplifiers will be presented. The fol- 
lowing section will review the design of 
tuned Class A amplifiers. 

This presentation is, by necessity, 
oversimplified. A more exhaustive treat- 
ment would carry us well beyond the 
scope of this volume. An attempt is 
made at justifying some rules of thumb 
which will be used later in the text. The 
reader who is not familiar with basic 
transistor concepts is urged to review a 
good basic treatment of the subject. The 
series of articles in QST by Stoffels is 
excellent. 1 

Consider first the simple amplifier 
shown in Fig. 7. This amplifier operates 
in Class A, which means that collector 

'Available in reprint form from ARRL for 

$ 1 . 

Basics of Transmitter Design 19 









current Hows during the entire drive 
cycle. 

First, we will review the biasing. The 
base is driven from a voltage source of 
4. Since we are using a silicon transistor, 
die emitter voltage will be less dian the 
base by about 0.7 volt, or 3.3 volts. The 
emitter current is 3.3 4- 500. or 6.6 mA. 
Since die collector current is virtually 
the same as the emitter current, the 
collector voltage is V cc 1 C R C = 8.7 
volts. 

This arithmetic is based on the as- 
sumption that the base is biased from a 
true voltage source. It's wise to confirm 
diis. The current in the base-voltage 
divider is 12 V 4- 15kf2or0.8 mA. If 
the 0 of our transistor is 1 00. the base 
current is l h = l c 4- 0 = 6.6 mA 4-100 = 
66 n A. Since the current in the divider 
is 10 times tliis value, our bias divider is 
indeed "stiff enough. 

These bias calculations describe die 
operadon of die transistor at dc. Our 
interest, however, is in the behavior of 
the circuit for an ac signal. For rf signals 
the emitter is essentially ac-groundcd 
through the emitter bypass capacitor. 
Recall that a capacitor is a device which 
has the characteristic that the impressed 
voltage cannot change instantaneously. 
Any rf signal that appears at the emitter 
of the transistor will be connected to 
the capacitor directly. Since the voltage 
at diis point cannot change instanta- 
neously (i.e.. at an rf rate), all ac parts 
of the emitter current flow through the 
capacitor rather than through the 500- 
ohm emitter resistor. Thus, we treat the 
amplifier as a grounded-emitter stage. 

The input resistance of a grounded 
emitter amplifier is approximated by 
R in = 250 4- l c , where the emitter 
current is in mA. The beta used in this 
equation is not the dc-current gain we 
used in the preceding bias discussion, 
but is the ac-current gain, which is well 
approximated by P A c = f T 4- f op . where 
f„ p is the operating frequency of the 
amplifier and /•/• is die usual gain- 
bandwidth product. If we use a tran- 


sistor with a 1 50-MHz //• at a frequency 
of 7 MM/., the ac beta is about 20. 
Hence, the input resistance of the ampli- 
fier is 75 ohms. The .01 -/iF capacitor in 
the input merely serves to block dc. 
That is, it allows a difference in dc 
voltage to exist between die amplifier 
input and the output of die previous 
stage, but offers essentially no imped- 
ance to the flow of rf currents. 

Let’s assume that die amplifier is 
driven with .01 volt (10 millivolts). The 
input current (rf only) will be = Ej„ 4- 
R in = .01 4- 75 = 0.133 mA. The 
collector signal current is then I c = 0/ ft 
= 20 (0.133 mA) = 2.66 mA. This 
current flows through a load resistor of 
500 ohms. Again, using nodiing but 
Ohm's Law, we see that the output 
voltage is 1.33. The small-signal voltage 
gain is 1 .33 4- .01 , or 133. 

What would happen if we increased 
the input drive from 10 mV toO.I volt? 
If we were to follow the foregoing 
analysis again, we would calculate an ac 
current of 26.6 mA in die collector. 
However, the dc current is only 6.6 mA. 
There is no way diat this can happen in 
a linear amplifier. On positive peaks of 
die input voltage, the collector voltage 
would be driven down until it was 
nearly at the voltage of the emitter. This 
condition is called saturation, and is 


typified by reduced current gain. On 
negative peaks of the input signal the 
collector current decreases from the dc 
level of 6.6 mA until it is zero. The 
current can't go negative in a transistor. 
At this point, there is no collector 
current flowing; hence, the output volt- 
age equals the supply voltage of 12. We 
see that our amplifier is clipping the 
output waveform on both positive and 
negative peaks. What can be done to 
avoid this distortion? There are three 
possible solutions. First, we can reduce 
die drive level. Second, we can increase 
the dc current (lowing in the stage while 
simultaneously reducing the output load 
resistance. Finally, we can introduce 
some negative feedback in the amplifier, 
thus bringing about a reduction in stage 
gain. 

Let's consider the feedback solution 
by analyzing the modified circuit of Fig. 
8, where emitter degeneration is intro- 
duced. First, we note that the dc resis- 
tances are the same as before. There- 
fore, die dc bias current has not 
changed from die previous 6.6 mA. 
Using this value we find the dc voltage 
across the capacitor, labeled V x in the 
schematic, to be 2.64 volts. This point is 
bypassed, so it cannot change in poten- 
tial when rf is applied. Assume now that 
a signal of 0.1 -volt peak is applied to the 
input (0.2-volt pk-pk). As die input 
voltage goes from 0 to 0.1 volt, the base 
voltage will increase by 0.1 volt. The 
emitter voltage will also increase and 
follow the base, going from die dc level 
of 3.3 to 3.4 volts. Noting that the V x 
point in the emitter circuit is bypassed, 
die emitter current will increase to an 
instantaneous value of (3.4 - 2.64) 4- 
100 = 7.6 mA. The collector current is 
essentially the same. Hence, the col- 
lector voltage will drop to V cc l c R c 
or 12 -7.6 (0.500) = 8.2 volts. But, the' 
dc voltage was 8.7 volts. Hence, die 
voltage change is 0.5 volt. The small- 
signal gain is now 0.5 volt peak 4- 0.1 
volt peak = 5. Note that die voltage gain 
is now the ratio of the collector load to 
die unbypassed part of die emitter 
resistor. 



20 Chapter 2 






Fig. 7 - Circuit of a simple Class A rf ampli- 
fier. 


We can extend this simple argument 
to show that the input resistance of the 
amplifier has increased. With an input 
signal of 0.1 volt peak, the collector 
current increased from 6.6 to 7.6 mA (1 
niA). Since the high-frequency beta of 
the transistor is 20, the base-current 
increase is 1 + 20 mA. The small-signal 
input resistance is given by 


fyii 


A V 
M 


0.1 


20 


X 10' 


= 2000 ohms 

(Eq. I) 


where the deltas signify a small change. 

In general, the input resistance of a 
transistor with emitter degeneration is 
l$R e , where R e is the unbypassed por- 
tion of the emitter resistance. 

By using emitter degeneration we 
have realized a number of goals. First, 
the distortion is removed, for the signals 
are significantly less than tire dc bias 
conditions in the amplifier. We have 
substantially increased the input resis- 
tance, making the amplifier much more 
effective as a buffer. Finally, we have 
realized a gain which is dependent upon 
resistor values, rather than upon tran- 
sistor characteristics. As a bonus the 
bandwidth of the amplifier will be 
significantly higher. 

In the form shown in Fig. 8 our 
amplifier is not especially useful, for the 
output is not connected to anything. All 
of the output power is being delivered 
to the 500-ohm collector value. Suppose 
we coupled the amplifier capacitively to 



Fig. 8 - Class A amplifier using emitter de- 
generation. 


a following stage with an input resis- 
tance of 500 ohms. The net load on the 
amplifier is now the parallel combina- 
tion of the two loads, or 250 ohms. 
With a reduced collector-load resistance 
the voltage gain has dropped to 2.5. 

The original voltage gain of 5 could 
be regained by replacing the collector 
resistor with a large inductor (i.c., an rf 
choke). An inductor is merely a com- 
ponent which resists any change in 
current flowing through it. (Note the 
analogy of an inductor to a capacitor. 
The L is to current what a capacitor is 
to voltage, with regard to circuit be- 
havior.) Willi an inductance supplying 
the dc current to the collector, but 
resisting any changes in current, all 
signal current must llow into the ex- 
ternal load. In this case, the load would 
be the 500-ohm input to the next stage. 

The simple amplifier could be modi- 
fied further by the addition of an 
emitter follower, as shown in Fig. 9. 
Since the emitter follower has a by- 
passed collector with the output signal 
taken from the emitter, we have a stage 
with unity voltage gain, but a very high 
input resistance. 

From earlier calculations, we found 
the dc collector voltage of Q1 lobe 8.7 
volts. Hence, the emitter potential of 
02 is 0.7 volts less than this, or 8.0 
volts. The current in the emitter of Q2 
is 20 mA. When a drive signal is applied 
to this two-stage amplifier, the emitter 
of 02 will follow the base, being 0.7 
volt lower. For positive-going excursions 
of the output, signal current will be 
supplied to the external load and to the 
400-ohm emitter resistor from Q2. On 
negative-going output excursions, how- 
ever. current is pulled out of the ex- 
ternal load resistance and is allowed to 
flow into the 400-ohm emitter resistor. 
In this case, the maximum current we 
can handle on the negative-going ex- 
cursions is 20 mA. peak. In general, the 
standing dc current in the follower must 
exceed the peak signal current that the 
emitter follower is required to deliver. 

Shunt Feedback 

The amplifiers just discussed use 
emitter degeneration, or series feedback. 
AnoUier type of feedback that is quite 
useful is shunt feedback, and is used 
typically with operational amplifiers. 
An example of an rf buffer amplifier 
using shunt feedback is shown in Fig. 
10 . 

Recall that a silicon transistor has an 
input offset of about 0.7 volt. That is, 
the base of a conducting transistor is 0.7 
volt above the emitter. Also, note that a 
common (grounded) emitter amplifier is 
an inverting amplifier. This means that 
an increase in base voltage leads to a 
decrease in collector voltage. With these 
ideas in mind, let's analyze the circuit of 
Fig. 10 



Fig. 9 — Class A amplifier followed by an 
emitter follower. 


With the emitter of Q1 grounded, 
die base potential must be 0.7 volt. This 
means that there will be 0.7 mA of 
current flowing through Rl. Where can 
this current come from? It certainly 
can’t be coming from the transistor — 
dc current flows into the base of an npn 
transistor rather than out of it. The 
current must be supplied by R2,a 5-kS2 
resistor. This resistor must also supply 
die base current to Ql. This current, as 
we will show, is small enough in com- 
parison with the 0.7 mA that we can 
ignore it. With 0.7 mA flowing in R2, 
we must see a voltage drop of 5 kf2 X 
0.7 m A, or 3.5 volts across R2. The 
output dc voltage at the emitter of Q2 
must therefore be 4.2. Again, noting 
that Q2 will have a 0.7-volt offset, the 
collector of Ql must be at 4.9 volts. 

Consider now an input signal applied 
to the amplifier of 0.1 volt peak (0.2- 
volt pk-pk or 70-mV mis). As the input 
signal increases from zero to +0.1 volt, 
the current through R3 will go from 
zero to some positive value. This current 
would tend to flow into the base of Ql . 
However, tiiis would cause the collector 
voltage of Ql, and hence the amplifier 
output voltage, to drop dramatically. 
This drop leads to a decrease in the 
current flowing in the feedback resistor, 
R2. The output voltage will drop until 
the net current flowing into the node at 
die base of Ql is just enough to keep 
the potential of die base at 0.7 volt. 



Fig. 10 — Rf buffer using shunt feedback. 


Basics of Transmitter Design 21 










That is, the effect of the input signal is 
to replace current flowing in the feed- 
back resistor with current flowing from 
the input resistor. The input voltage is 
maintained at 0.7 volt in this amplifier 
when the output drops from the dc 
value of 4.2 volts to 3.7 volts. The 
voltage gain is 

G = ^£i!'= (3.7 — 4.2) _ , 

v AK,„ 0.1-0 

(Eq. 2) 

The minus sign indicates that the ampli- 
fier is inverting. 

Note that the gain depends upon the 
resistors and not upon transistor charac- 
teristics. That is G v = R2 -4- R3 = -R fb 
■=■ Rin ■ Also, the potential at the base of 
Q1 has been maintained essentially at 
0.7 volt because of the feedback. This 
means that the input resistance looking 
into the base is virtually zero. The input 
resistance of a shunt fed-back amplifier 
approximates the value of the input 
resistor (R3). This well-defined input- 
resistance characteristic is independent 
of the load effects at the output, 
making such an amplifier ideal for buf- 
fering and isolation purposes. 

Consider, finally, what would hap- 
pen at the output if we were to increase 
the load, or ask the follower for more 
output current. This might correspond 
to keying a following stage. Owing to 
the feedback, the output voltage will 
adjust until the input offset of 0.7 volt 
at Q1 is maintained, with Q2 delivering 
whatever current is needed to do this. 
Hence, output impedance is reduced by 
shunt feedback. 

The examples discussed here have 
demonstrated the use of series and 
shunt feedback. Admittedly, the anal- 
ysis was highly simplified. What is, 
perhaps, surprising is that in many cases 
the simplistic analysis presented is more 
than adequate for design purposes. 
Many of the buffer amplifiers used in 
the projects described later were de- 
signed by using these methods rather 
than a more elegant approach. 

Irrespective of the accuracy of the 
analysis, we can certainly use the results 
qualitatively to improve our intuition 



Fig. 11 — Class A amplifier with emitter de- 
generation and a tuned collector circuit. 


about feedback circuits. This will guide 
us in our experimental efforts. Negative 
feedback, (series or shunt) will always 
decrease the amplifier gain. It will also 
increase the bandwidth. Series feedback 
will have the effect of raising the input 
impedance while shunt feedback will 
decrease both the input and output 
impedances. Feedback amplifiers will be 
discussed in more detail in the chapter 
on ssb methods. 

Tuned Buffer Amplifiers 

The previous circuits used resistive 
loads. However, most buffer amplifiers 
will be tuned. The use of resonant 
circuits improves the performance in a 
number of ways. Higher gain is possible, 
selectivity is introduced into the re- 
sponse of the circuit, and finally, higher 
power outputs are possible, since a high 
standing current can be used while 
maintaining a high collector voltage. 

In this section, we will extend the 
designs described earlier to the case of 
tuned output loads. The rudimentary 
details of how a tuned circuit is treated 
analytically and how it is used for 
impedance matching will be presented. 
The first example is shown in Fig. 11. 
This circuit is nearly identical with that 
discussed in Fig. 8 where emitter degen- 
eration was introduced. 

Before considering the behavior of 
the amplifier of Fig. 11, we should 
review the nature of a simple tuned 
circuit. A toroidal inductor has been 
used. Toroids have distinct advantages 
for the experimenter. First, the mag- 
netic field of a toroid is contained 
almost completely within the core. As a 
result, minimal magnetic energy from 
the tuned circuit will couple into other 
parts of the circuit to cause instability. 
This is not the case for a solenoidal 
inductor. The second advantage is that 
the inductance of a toroid is described 
by a simple and quite accurate equation 
which can simplify things for the de- 
signer. Knowing the number of turns 
(A) on a toroid the inductance is L = 
KN 1 , where K is a proportionality 
constant. For the Amidon T-50-2 core 
used in our design example, the con- 
stant is 5 nanohenrys (nH) per turn 
squared. Thus, the inductance of a 
30-turn winding on this core is L = 5 
nH/r J (30/) 2 = 4500 nH = 4.5 pH. 
Data are presented in the appendix for a 
number of popular toroid cores avail- 
able to the amateur. 

This amplifier will be operated at 7 
MHz. The capacitance required to res- 
onate the inductor on 40 meters is given 
as C = 1 t (27 if) 2 L. In this case, C = 
115 X 10' 13 farad, or 115 pF. In 
practice, one might use a 180-pF mica- 
compression trimmer capacitor. Alter- 
natively, a low -capacitance variable 
could be paralleled with a fixed-value 
mica capacitor. 



Fig. 12 - Schematic representation of circuit 
losses. 


We have now described the super- 
ficial details of our circuit by specifying 
the inductance and capacitance. How- 
ever, additional information is needed 
for circuit analysis. 

If a quantity of energy is injected 
into a tuned circuit, that energy will 
remain stored for a reasonable time. A 
voltage across the capacitor will cause 
current to flow in the inductor. How- 
ever, current flowing in the inductor 
will lead to a voltage being developed 
across the capacitor. If there were no 
loss elements in the tuned circuit the 
energy would remain stored forever. 
Any real tuned circuit, however, does 
have losses. In the hf region and at 
lower frequencies the predominant loss- 
es are associated with the inductor. 

The presence of losses leads us to 
define a pertinent term, Q, which is a 
figure of merit for a resonator. Formal- 
ly. Q is defined as the total energy 
stored in a tuned circuit, divided by the 
energy lost in one cycle of oscillation. It 
may be shown mathematically that this 
Q is also related to the bandwidth by Q 
= f + Af, where Af is the 3-dB band- 
width. 

For circuit applications, still another 
means is needed to model the losses of a 
tuned circuit. This can be done by 
assuming that our real and lossy tuned 
circuit is replaced by a perfect one with 
a resistor, either in series or parallel with 
the inductor. Using this representation 
it may be shown that these resistors are 
related to the Q and the inductance by 
Q - R p -=• 2 irfL = 2nfL -r R s . Schematics 
showing these loss resistances are pre- 
sented in Fig. 12. 

If the tuned circuit has no other 
elements attached to load it, the Q 
realized is called the unloaded Q, or Q u . 
On the other hand, if energy is ex- 
tracted from the LC combination and 
used for some other purpose, the re- 
sulting Q is the loaded Q. 

For the T-50-2 toroid used in our 
amplifier example, the typical Q at 7 
MHz is 150. ( Q is a dimensionless 
number.) Since the inductance is 4.5 
MH, the parallel-equivalent loss resis- 
tance, R p , is given as R p = Q2nfL = 
29.7 kft. If we were to shunt the tuned 


22 Chapter 2 





Fig. 13 - Output coupling from a Class A amplifier using a toroidal transformer in the col- 
lector circuit. 


circuit with an external 5-kf2 resistor 
the net parallel-equivalent resistance 
across the coil would be 4.28 kO. Hence 
the loaded Q would be 4.28 k v 2 vfL = 
21.6. The loaded Q is always less than 
the unloaded Q. 

How do we treat this parallel com- 
bination of an inductor, a capacitor and 
a resistor when they appear in a circuit? 
In general, it would be necessary to 
consider the parallel combination of all 
of the impedances in order to arrive at a 
suitable equivalent impedance for use in 
an analysis. However, at resonance the 
case is simplified considerably, for the 
parallel capacitor and the inductor have 
the effect of canceling each other, in 
terms of reactance leaving the parallel 
resistor as our equivalent impedance. 
Indeed, this is the definition of reso- 
nance. 

We are now in a position to return to 
the original amplifier of Fig. 1 1 and tp 
calculate its gain. At resonance, the 
tuned circuit appears to be a 29.7 -kO 
resistor. The voltage gain of the circuit 
is 29,700 -r 1 00, or 297. This gain is 
extremely high. In fact, it is so high that 
the chances of instability are very good. 
Ignoring this potential problem, we note 
that this high gain is obtained while 
keeping 12 volts of bias on the col- 
lector, and several mA of current 
flowing. This could not be realized 
without a tuned circuit. 

In order to extract some energy 
from the output of the amplifier, as- 
sume that a 5 -turn link is wound over 
the toroidal inductor (see Fig. 13). An 
asset of toroids is that almost unity 
coupling is provided between various 
windings on the core. It was this unity 
coupling that led to the simple N 1 
inductance formula described earlier. 
Another feature is that impedances 
terminating one winding are trans- 
formed to the other winding according 
to the square of the ratio of the turns. 
Hence, if a 50-ohm resistor is placed 
across the 5-turn link, this has the same 
effect as a parallel load resistor across 
the tuned circuit where: R L = (30 4- 5) 2 
X 50 = 1800 ohms. This external load 


appears in parallel with the 29.7 -kf2 
resistor which represents the core losses, 
resulting in a net load of about 1 .7 kS2. 
With this load, the voltage gain of the 
circuit (collector voltage divided by base 
voltage) is 17, a high but probably 
stable value. Further, the loaded Q of 
the resonator is Q L = R L -r 2itfL = 
1 ,700 t 198 = 8.5, where R L is the net 
load resistance. The loaded bandwidth 
will then be about 800 kHz. 

Assume now that the amplifier is 
excited by a 0.1 -volt peak signal at the 
input. The ac signal at the collector will 
be 1.7 volt, peak. The rf collector 
current is just 1 .7 V -r 1 .7 kf2, or 1 mA. 
Since this is well below the dc current 
standing in the stage, the linearity 
should be excellent. 

Since the turns ratio on the tuned 
transformer is 6:1, the voltage across 
the 50-ohm load resistor is just 1/6 the 
collector voltage, or 283-mV peak. 

If this amplifier were driven from a 
low-impedance source, the net voltage 
grin would be only 2.8, and we would 
not consider this to be much of an 
amplifier. However, the buffering is 
quite good since the input resistance 
was 2 kfi (see previous section). If the 
input to this amplifier were impedance 
matched, the gain would be a little over 
25 dB - a very respectable value. 

Power Output 

It is interesting to calculate the 
maximum power output which can be 


obtained from this stage while main- 
taining Class A operating conditions. In 
the example outlined above, we saw an 
ac signal on the collector of 1.7 -volts 
peak. The dc collector potential was 12 
volts. Hence, the instantaneous voltage 
on the collector would vary from 10.3 
to 13.7 volts at a 7-MHz rate. Note that 
the collector potential exceeds the 
+1 2-volt dc bias. 

The emitter dc voltage was 3.3 volts. 
As an approximation we will neglect the 
fact that the emitter is not totally 
bypassed. The maximum signal voltage 
we could expect to see on the collector 
would be (12 - 3.3) = 8.7 volts peak. 
That would be the signal which would 
cause the transistor to just go into 
saturation on negative peaks. The posi- 
tive voltage peak would be (12 + 8.7) or 
20.7 -volts peak. The pk-pk signal is just 
twice 8.7, or 17.4 volts. 

If our amplifier is to stay linear 
(barely) during this voltage excursion, 
the current must be fluctuating from 
zero to twice the dc value of 6.6 mA. 
Now we ask what the proper load 
resistance would be to obtain these 
swings in voltage and current simulta- 
neously. This is given again by Ohm’s 
Law, as R L = (8.7-V peak) -5- (6.6-mA 
peak) = 1.32 k£2. If we increased our 
link from 5 to 6 turns, the load pre- 
sented to the collector would be 1.25 
kf2, a close approximation. With this 
load the maximum power output will be 
given as l 2 R = (6.6 X 10 -3 ) 2 X 1.250 
kf2 = 54-mW peak, or 27-mW rms. 

In this example we will not, in 
practice, be able to obtain quite this 
much output. This is because on 
negative-going output peaks, when the 
transistor approaches saturation, the 
emitter voltage will rise above the 
3.3-volt dc level. On the other hand, if 
this amplifier were slightly overdriven, 
die dc collector current would rise 
above die 6.6-mA bias level and some 
additional power output could be ob- 
tained. This nonlinear mode of oper- 
ation is often used in cw applications. 

In most linear applications it is 
desirable to maintain Class A operating 
conditions where the stage current does 
not fluctuate with drive level. While ssb 



Basics of Transmitter Design 23 




is the obvious application, in some cases 
this is also advisable during cw oper- 
ation. The reason is that a linear ampli- 
fier tends to maintain the selectivity 
inherent in the resonators. On the other 
hand, if the amplifier is allowed to 
saturate additional loading occurs across 
the tuned circuits, and that decreases 
the selectivity. That can have the effect 
of increasing spurious output, especially 
when the stage is driven by a mixer or 
frequency multiplier. 

Load Resistance 

Now that we have analyzed an ex- 
ample we are in a better position to ask 
a more general question. That is, what 
load resistance should be used for a 
specific power output? If we consider 
only amplifiers which have bypassed 
emitters, the load required is that resis- 
tance which will allow the collector to 
fluctuate with a peak voltage excursion 
equal to the difference between the 
supply, V cc , and the emitter voltage. 


V e . For a given resistance the power 
delivered to that load is V rms 2 + R, or 
Vpeak 2 * 2/?. Solving this for the load 
resistance, we have R L = ( V cc - V e ) 2 -f 
2 P„. The Class A amplifier should be 
biased to a current equal to the peak 
signal current, which is I dc = 2 P„ + R/,. 

Although we have been discussing 
Class A amplifiers predominantly, the 
expression for load resistance is quite 
general and applies as well to Class C 
amplifiers. In the typical Class C power 
amplifier, the emitter is at dc ground, 
leading to the well-known expression 
/?/. = V ce 2 T IP 0 . 

The last two expressions may be 
combined to show that the maximum 
efficiency of a tuned Class A amplifier is 
50 percent. In practice, efficiencies near 
30 percent are more common, especially 
if good linearity is desired, as would be 
the case with ssb. 

It is desirable when seeking selectiv- 
ity to use circuits other than a single 
LC combination in the output of a Class 
A stage. An example might be the first 
buffer amplifier following a mixer in a 
heterodyne exciter of the kind that 
might be used in a ssb transmitter. 


Although rather complex, the same 
basic principles apply. The gain of the 
amplifier is determined by the imped- 
ance “seen” when looking into the 
input of the more complex filter. 

In some cases a filter may require a 
given termination at its input in order to 
provide the desired selectivity. In this 
case it may be required to resistively 
terminate the collector of an amplifier 
in order to present the proper load to 
the following filter. The gain of such an 
amplifier will depend upon the resistor 
and filter characteristics. This situation 
is illustrated in Fig. 14. The appendix 
contains a catalog of two- and three- 
section filters for the amateur bands. 
They are suitable for such applications. 

Earlier, it was shown that shunt 
feedback in an amplifier has the effect 
of decreasing the output resistance of 
that circuit. Therefore, by careful use of 
feedback the output impedance of an 
amplifier can be adjusted to provide the 
proper input termination for a multi- 


pole filter. Designs of this kind are 
practical and will be covered in tire ssb 
chapter. 

The Medium-Power Class C 
Amplifier 

When high output power is desired 
for the final stage of a QRP transmitter, 
or from the driver in a medium-power 
cw transmitter, a Class C amplifier is 
usually chosen. While these stages lack 
the envelope linearity needed for ssb, 
they offer high power gain, high power 
output and good efficiency. In this 
section, amplifiers with an output up to 
2 watts will be considered. 

A Class C amplifier is defined as one 
where collector (or plate) current flows 
for less than half of the drive cycle. The 
normal transistor amplifier operated 
with no reverse bias on the base is 
actually a Class C amplifier, since there 
is in effect a built-in bias in the tran- 
sistor. That is, the base voltage must 
exceed 0.7 volt positive before con- 
duction occurs. 

At this point, we will shift gears 
slightly, away from a simple analytic 
treatment toward a more empirical ap- 


proach to the design problems. In gen- 
eral, the small-signal approximations 
used in the previous text are not too 
accurate in die description of Class C 
amplifiers. Nonetheless, we can extend 
our previous understanding to describe 
qualitatively a high-power Class C ampli- 
fier. For example, the gain of such a 
stage is still determined by the high- 
frequency beta of the transistor, which 
is in turn, a function of the f T of the 
device. The maximum output power 
will be limited by the load impedance 
we present to the collector. As the dc 
current level is increased, and hence, the 
power level of the stage, the input 
resistance decreases. 

Shown in Fig. 15 is a Class C 
amplifier coupled by a link from an 
earlier stage. 

Starting at the input, the first con- 
sideration is to determine the turns ratio 
of the input transformer. If the base of 
the power amplifier were a simple resis- 
tive input, as is essentially the case with 
a Class A stage, the turns ratio would be 
determined by the simple impedance- 
matching criterion outlined earlier. 
However, the input to the Class C 
amplifier is not, in the general case, a 
pure resistance. At low frequencies a 
better model for the input would be a 
silicon diode with some series resistance. 
Unlike the usual silicon diode, however, 
the one used in our model (representing 
tlie power transistor) will have a low 
reverse-breakdown voltage. Typical 
values will be 3 to 5 volts. The input 
link must be chosen to deliver current 
to the base on positive peaks of the 
driving voltage. However, the open- 
circuit voltage from die link must be 
low enough that the reverse breakdown 
of the diode is not exceeded. The driver 
should have a power output consistent 
with the expected power gain of the 
Class C stage. That is, if an output of 1 
watt is desired, and we expect a gain of 
16 dB in die Class C amplifier, we 
should have 25 mW available from the 
previous stage. 

If the reverse breakdown of die 
base-emitter junction of the power 
amplifier is exceeded, die result is not 
an instantaneous catastrophe: The tran- 
sistor does not go up in smoke. However, 
the long-term result is just as devasta- 
dng. Prolonged operation with the input 
diode being switched into breakdown 
will lead to a deterioradon in the 
current gain of die transistor. Hence, 
the power output will continually drop 
off. 

This effect can be observed easily 
with small-signal transistors operating at 
very low frequencies. A simple experi- 
ment can be done to demonstrate it. 
Start with an inexpensive plastic tran- 
sistor, for this is a destructive test. 
Measure the dc beta of the transistor at 
a collector current of, say. 10 mA. Then 



24 Chapter 2 



apply a reverse bias to the emitter-base 
junction with current limiting to keep 
the “Zener” current at around 10 mA. 
Operate the transistor for about an hour 
in this manner. Then, again measure the 
dc beta. A degradation will usually be 
noted. Low-level transistors are often 
used as Zener-diode substitutes by oper- 
ating the e-b junction as outlined. This 
practice is generally fine. However, once 
used as a Zener diode, the device should 
be retired from service as a transistor. 

It is generally more difficult to 
observe this phenomenon at high fre- 
quencies. It is straightforward, however, 
if the experimenter is fortunate enough 
to have a high-frequency oscilloscope in 
his shop. This problem is generally 
limited to transmitters on the lower- 
frequency amateur bands, usually at and 
below 7 MHz. The reverse base break- 
down is prevented by choosing carefully 
the turns ratio in the driving circuitry 
and by keeping the value of the shunt 
resistor at the base fairly low. A resistor 
in series with the base should be 
avoided. 

Returning to Fig. 15, the resistor 
shunting the base serves two functions. 
First, it provides a load for the driver 
during the negative voltage excursions 
of the driving signal, and hence, pre- 
vents the reverse breakdown from oc- 
curring, as outlined. Second, it absorbs 
some drive energy that might otherwise 
find its way into the base. Since part of 
this energy could result from feedback 
in the amplifier as well as from the 
driver, the resistor decreases stage gain 
and tends to stabilize the amplifier. If 
instability is ever noted in a Class C 
stage, the first thing to do in order to 
“tame the beast” is to decrease the 
ohmic value of this shunting resistor. 

As a rule of thumb with amplifiers 
operating from 12- to 15-volt supplies, 
the driving link is approximately 1/10 
the number of turns used in the primary 
of the driver transformer. Typical values 
for the base resistor in 1- to 2-watt 
amplifiers is 18 to 100 ohms. 

When the operating frequency of the 
amplifier is increased to roughly a tenth 
of the f T of the transistor (or higher), 
the input of the transistor ceases to look 
like the simple diode model outlined 
previously. Charge-storage effects within 
the transistor make the input appear 
much more like a resistive input shunted 
with a capacitance. Modern transistors 
designed specifically for rf power appli- 
cations have the input resistance and 
capacitance specified by the manu- 
facturers. As odd as it may seem the 
reduced power gain and more stable 
input characteristics which occur at high 
frequencies often make it much easier 
to build amplifiers which operate 
toward the high end of the spectrum. 
(We’ve encountered many more stability 
problems with 1-watt amplifiers on the 


160-meter band than we have seen at 
144 MHz!) 

Output Circuit 

Designing the output circuit is simi- 
lar to tire procedure described for the 
Class A amplifier in the previous 
section. With no drive power present, 
the base of the Class C amplifier is at 
ground and the transistor draws virtual- 
ly no current. Only when drive is 
applied does any collector current flow. 
This current in the collector will cause 
tire voltage at tire collector to depart 
from the quiescent value of V cc . If we 
assume that the collector voltage varies 
from 0 to twice the V cc level while 
delivering tire desired output power, the 
load needed at the collector is given by 
tire familiar relation /?/. = V cc ? + 2P„. 

There are a number of networks 
which can be designed to transform a 
50-ohm termination to any desired prac- 
tical resistance. These are outlined in 
chapter 4. For stages operating at power 
levels such that R L is 50 ohms or 



Fig. 16 - Example of a link-coupled output 
network. 


higher, link coupled output networks 
can often be used satisfactorily. For 
higher powers other networks arc rec- 
ommended. 

As an example of a link-coupled 
output network consider the stage 
shown in Fig. 16. We will design for an 
output power of 1/2 watt at 14 MHz, 
and we will use a transformer with 15 
turns as the major resonant winding on 
an Amidon T-50-2 toroid core. The 
inductance is ( 1 5 ) 2 X 5 nH-f' 2 .or 1.13 
pH. This will resonate at 20 meters with 
a capacitance of 115 pF. We will design 
for a loaded Q of 6 and a supply voltage 
of 12. 

The inductive reactance of the coil is 
99.4 ohms. With a Qt. of 6, we thus 
want the 50-ohm link to present a 
parallel resistance across the coil of 
QilufL, or 596 ohms. Noting that 
impedances transform in proportion to 
the square of the turns ratio, we see that 
the output link should have 0.29 the 
number of turns of the main winding 
(4.35 turns). We will use 4 turns. 

With a 12-volt supply, the load we 
want to present to the transistor is V cc 2 


-r 2P a = 144 ohms. The turns ratio 
between thi s winding and the 50-ohm 
winding is \/144 -5- 50 = 1 .7. Since the 
50-ohm winding has 4 turns, we calcu- 
late that the transistor winding should 
be 6.8 turns. A 6- or 7-turn link will do 
the job. The parallel resistance repre- 
senting the unloaded Q of the coil has 
been neglected since the loaded Q of 6 
is much less than the inductor unloaded 

Q- 

Once a suitable network is designed 
and implemented, the maximum power 
output is defined. To realize this output 
the stage must be driven adequately. If 
the drive is less than that required for 
full power output, the collector voltage 
will not swing from ground to twice 
y cc . but something less, centered 
around Vcc- Such operation is typical 
for linear amplifiers used for ssb applica- 
tions. However, for cw use, the ampli- 
fier is usually driven to full output since 
this results in maximum efficiency. 

Components 

The collector rf choke is a com- 
ponent which is often treated too 
casually. The choke should have a low 
dc resistance, for any IR drop in the 
choke will subtract from the available 
supply voltage. The inductance of the 
choke should not be excessive. Too 
much inductance will cause resonances 
to exist with the capacitors in the 
output network which are much lower 
than the output design frequency. Since 
the typical transistor has a gain which is 
increasing dramatically at lower fre- 
quencies, these resonances can lead to 
instabilities. A reasonable rule of thumb 
is that the output rf choke should have 
a reactance at the operating frequency 
which is between 5 and 10 times R L . 
An additional (and wise) precaution is 
to parallel the usual 0.1 -pF bypass 
capacitor with an electrolytic capacitor 
of around 10 /zF. 

The general criteria for selecting 
transistors for amplifiers of this kind are 
/r, breakdown voltage, power dissipa- 
tion and maximum current. The f T 
should be well above the operating 
frequency: however, not by too much. 
It is sometimes quite difficult to use vhf 
power transistors on the lower hf bands 
due to the tremendous gain available, 
which causes instability problems. The 
collector breakdown voltage should be 
twice the supply to be used, although 
this rule can sometimes be violated 
because the transistor is not conducting 
during the period when the highest 
collector voltages are present. In gen- 
eral, the power dissipation of the tran- 
sistor should be at least as high as the 
output power desired. This also implies 
that a heat sink may be necessary if it is 
needed to realize the dissipation rating. 
The maximum collector-current capabil- 
ity of the transistor should be at least 


Basics of Transmitter Design 25 




Fig. 17 — Schematic diagram of the universal QRP transmitter. Resistors are 1 /2-watt compo- 
sition. Cl is a trimmer capacitor. C3 and C4 are silver-mica capacitors. Remaining capacitors 
are disk ceramic. 50 volts or greater. See text for Q1 , Q2 types. Component values not on 'he 
diagram are listed in Table 1. 


twice the dc current expected. 

The efficiencies of Class C amplifiers 
in the 1- to 2-watt category vary con- 
siderably. but are usually around 60 
percent. Efficiencies of over 75 percent 
are not uncommon. If the efficiency is 
under 50 percent, a better output tran- 
sistor might be in order. 

A Universal QRP Transmitter 

The ideas outlined previously can be 
applied to the design of a simple two- 
stage transmitter for the hf or 160- 
meter bands. Although the seasoned 
QRP operator may scoff at a non-VFO 
transmitter, the use of crystal control 
can lead to simplicity as well as an 
uncompromisingly clean signal. The de- 
sign lends itself well to the later ad- 
dition of a VFO. 

The essential details of the trans- 


mitter are shown in Fig. 17. Only a few 
of the component values are specified 
on the schematic. The rest vary from 
band to band and are summarized in 
Table 1 . 

The transmitter is near the ultimate 
in simplicity, consisting of a crystal- 
controlled oscillator driving a single- 
stage power amplifier. The crystal oscil- 
lator is keyed in all versions but the 
10-meter one. In the output stage a pi 
network is used to match the 50-ohm 
antenna to the collector of the ampli- 
fier. In this case the word “match" is a 
bit of a misnomer, for the network 
shown presents no impedance trans- 
formation. When the output is termin- 
ated in 50 ohms, a load resistance of 50 
ohms is presented to the collector of the 
final. However, the network acts as a 
low-pass filter to attenuate harmonics. 


The maximum power output which can 
be expected is about 1 .44 watts when 
using a 12-volt supply. Indeed, the 
measured output is just about 1-1/2 
watts on all bands except 10 meters, 
where the power is still over 1 watt. 

In the schematic, a capacitor, C5, is 
shown from the base of the oscillator to 
the emitter. This capacitor is used only 
on the 160- and 80-meter bands. 

On the bands up to 14 MHz, 
fundamental-mode crystals were used. 
In the test units, HC-6 type plated 
crystals were chosen. Several surplus 
FT-243 style 7 -MHz crystals were used 
in the 40-meter unit. They all oscillated 
readily and keyed well. 

On the 10- and 15-meter bands, 
third-overtone crystals were required. 
Since most 40-meter crystals will oscil- 
late readily on their third overtone, the 
7-MHz crystals also operate well in the 
15-meter transmitter. When FT-243 
crystals were used, the 21-MHz output 
was excellent, as was the keying. 

The reader will note that only one 
design is presented for both the 10- and 
the 15-meter bands. The circuit func- 
tions well on both of the bands by 
merely re tuning Cl , the capacitor which 
resonates the crystal oscillator. 

A minor problem was observed with 
the 10-meter design. It was found that 
there was a sligfit chirp when the oscil- 
lator was keyed. This was eliminated by 
rebiasing the stage for reduced output, 
but the drive to the final was then 
inadequate. Best 10-meter operation of 
this rig resulted from keying only the 
final, as shown in Fig. 18. Here, a pnp 
transistor is used as a switch, allowing 
the key to remain at ground potential. 
An even better solution would be to 
modify the design with a keyed Class A 
buffer between the oscillator and the 
output amplifier. This approach was 
taken in a 6-meter transmitter described 
at the end of this chapter. 

The number of transistors which can 


Table 1 



Cl 

C2 

C3 

C4 

C5 

LI 

L2 

L3 

R1 

RFC 

160 M 

400 pF 
MAX 

1800 pF 

1800 pF 

1800 pF 

360 pF 

73t 

No. 28 
T-50-2 

8t 

30t 

No. 26 
T-50-2 

1812 

50 pH 

80 M 

400 pF 
MAX 

100 pF 

750 pF 

750 pF 

200 pF 

43t 

No. 26 
T-50-2 

5t 

21 1 

No. 22 
T-50-2 

3912 

25 pH 

40 M 

180 pF 
MAX 

100 pF 

470 pF 

470 pF 


35t 

No. 26 
T-50-2 

4t 

14t 

No. 22 
T-50-2 

3912 

15 pH 

20 M 

60 pF 
MAX 

33 pF 

210 pF 

21 0 pF 

“ 

27 1 

No. 24 
T-50-6 

3t 

12t 

No. 22 
T-50-6 

4712 

15 pH 

15/10 M 

60 pF 
MAX 

33 pF 

105 pF 

130 pF 


17t 

No. 24 
T-50-6 

3t 

9t 

No. 22 
T-50-6 

4712 

15 pH 


26 Chapter 2 




Fig. 1 8 - Modification of the keying circuit for the 28-MHz version of the QRP transmitter. 


be used in this design is nearly endless 
and is growing daily. In test units built, 
the oscillator was either a 2N2222A or a 
2N3904. These devices are inexpensive 
and readily available. Other good candi- 
dates would be the 2N4124, 2N3641, 
2N3563, 2N3866, 2N3692 or 2N706, 
to mention only a few. 

In all of the units built, the final 
amplifier was a Motorola 2N5859. This 
is a TO-5 device similar to the RCA 
2N5189. The differences between the 
two are minimal. The 2N5859 is per- 
haps a bit “hotter,” with the 2N5189 
being slightly more rugged. A small 
smokestack type of heat sink was used 
on the output transistor in all units. 

When 2N5859s were used, they ap- 
peared to operate reliably when the 
transmitter was terminated properly in a 
50-ohm antenna with a VSWR of under 
2:1. However, the potentially destruc- 
tive testing procedure to be described in 
the following section showed that the 
transistors would not survive a severe 
mismatch. A Motorola 2N3553 was 
substituted in several of the units and 


the power output was the same. The 
output transistor could not be destroy- 
ed under the worst mismatch that could 
be found. Additionally, the higher 
power dissipation and breakdown volt- 
age rating* of the ’3553 allow the 
transmitter to be operated at up to 28 
volts, a level at which several watts of 
output power can be obtained. In this 
case, careful heat sinking is required. 
While this transistor is specified as a vhf 
power device, the cost is only S2.30 in 
single lots. 

Shown in Fig. 19 is a printed-circuit 
layout for the universal transmitter. 
This board is single sided and is only 2 
X 3 inches. The builder may want to 
make the board slightly larger if it is to 
be used on 160 or 80 meters, where the 
components are bigger. Likewise, the 
10-meter version could be reduced in 
size, if desired. 

Tuning of this family of transmitters 
is straightforward. After the unit is built 
and carefully inspected to ensure that 
the parts are in the proper slots, a 
dummy load, pcwer supply and crystal 


are connected. Some means of moni- 
toring the transmitter output is needed. 
Such a QRP power meter is described in 
a later chapter, although a suitable 
substitute would be a 51 -ohm, 1-watt 
resistor as the output termination with a 
VTVM/rf-probe combination for 
measuring output. Ideally the power 
supply should be current limited to 
around 0.25 A. With the power on and 
the key closed, the oscillator tank is 
tuned for maximum power output. The 
keying is monitored in the station re- 
ceiver, just to be sure it’s clean. That's 
it! Debugging, should problems occur, is 
covered in the next section. 

Fig. 20 shows a photograph of the 
160-metcr board. Shown also is a box 
which contains the 20-meter version. 
The packaged unit contains a slide 
switch which transfers the antenna and 
the 12-volt supply to the final stage 
during transmit intervals. The rear of 
the box contains a pair of bnc coax 
connectors for the antenna and receiver 
as well as banana jacks for the dc power 
input. Dc voltage is always applied to 
the crystal oscillator. This allows the 
operating frequency to be spotted by 
merely hitting the key. 

The 20-meter version was used for a 
couple of months of casual operation in 
the spring of 1974 by W7IYW. Al- 
though only one crystal was available, 
contacts were made with KH6, UA0, J A, 
ZL, VK, KX6 and G as well as with a 
few stateside amateurs. The 3-element 
Yagi antenna (at 80 feet) and an excel- 
lent location helped. Similar results can 
be expected with a dipole or ground 
plane vertical in a typical location, 
although the contacts will not come as 
easily, and the reports are sure to be 
down by a couple of S units. 

Construction Methods, Testing 
Techniques and “Bandaids” 

In the earlier sections of this chap- 
ter, the discussion has been rather basic 
with emphasis on the fundamentals. 
One design example was presented in 
the preceding section, but not very 
much has been said about construction 
and debugging of solid-state circuits. 
There are a few rules which make a 
profound difference in the performance 
obtained. 

Once a design has been transferred 
to a hardware form, it still may not 
function exactly as originally envisioned 
by the designer. Indeed, it is only in rare 
cases that debugging of some sort is not 
required. Some problems will be covered 
in this section. The reader is refer- 
enced to a QST paper on this subject 
which is especially good. 2 

As one reads the various amateur 
publications, he soon realizes that 

1 DeMaw, "How to Tame a Solid-State Trans 

milter," QST for Nov. 1971. 




\ 

RFC 

1 




-lpF 


.05»iF 


L3 


RF OUT 


GNO 


Fig. 19 - Scale layout of the universal QRP transmitter pc board. 


Basics of Transmitter Design 


27 






Fig. 20 — Photograph of the assembled QRP transmitter for 20 meters. At the left is a 160- 
meter version. 


almost all of the equipment built by 
today’s amateur experimenter is fabri- 
cated on etched circuit boards. One 
might assume that this is done merely to 
allow easy duplication and repeatability 
of performance and to impart a pleasing 
appearance. After all, that’s what the 
professionals do. In reality there is a bit 
more to it than this, especially when rf 
circuitry is concerned. A proper pc 
layout has the major advantage of pre- 
senting a low impedance return to 
ground wherever it is desired. This 
characteristic provides ample justifica- 
tion for using pc-board methods when 
building rf circuits! 

The amateur magazines and refer- 
ence books contain data for layout and 
etching of pc boards. These will not be 
repeated in detail here, for the methods 
are straightforward and easy to apply in 
the home. The builder is, however, 
cautioned to keep the basic goal of 
proper grounding in mind When de- 
signing a layout, even if it means that 
some of the aesthetic qualities of the 
board might be sacrificed. 

The best way to ensure a clean 
ground plane for an rf circuit is to use 
double-sided board (copper on both 
sides). This may present a minor 
problem to those who frequent only the 
local outlets where single-sided board is 
sold. However, when surplus outlets are 
investigated one finds that double-sided 
board is the rule rather titan the ex- 
ception. If modern electronic equip- 
ment is studied by the reader, he will 
notice that single-sided boards are 
seldom used. The norm these days for 
densely packed circuits is multi-layer 
boards, often containing up to 6 or even 
more individual layers. 

The easiest way for the amateur to 
use double-sided board, especially if 
one-of-a-kind boards are being built, is 
to use one foil for nothing but the 
ground plane. All soldering pads and 
runs are on the other side of the board. 
Once the side containing the “meat” of 


the circuit is etched and washed, and 
tire resist is removed, the holes are 
drilled in the board. Then, a large drill is 
used as a counter-sink to remove the 
copper from around all of the holes on 
the solid-foil side of the board. Then the 
components are inserted, being 
mounted on the ground-plane side of 
the board, and soldering can commence. 
Whenever a connection to ground is 
desired, the component is soldered di- 
rectly to the ground foil with the 
shortest possible lead length on the part. 
Numerous examples are shown in the 
photos throughout the book. 

All of the etched boards used in the 
illustrative examples of this book were 
built in the home lab. The resist ma- 
terials used were small pads or strips of 
Scotch brand electrical tape, or masking 
tape. In some cases a resist-ink pen was 
used. Ferric chloride was used as the 
etchant. The resist material used to 
protect the ground plane during etching 
was a layer of enamel spray paint, or 
full-width strips of masking tape or 
Scotch electrical tape. 

A series of QST articles featured 
circuit boards which are not etched. 3 
Instead, a hacksaw was used to cut a 
series of shallow grooves in the board, 
through the foil. This leaves a checker- 
board pattern of copper islands to 
which components may be soldered. 
Some of the equipment described in 
later chapters was built using a modifi- 
cation of this method. Double sided 
board was chosen, and a hacksaw was 
used to create the matrix of islands. 
However, the components were mount- 
ed on the groundplane side of the 
board. Holes were drilled in exactly the 
same way as with an etched board. If 
the copper islands are kept fairly small, 
tire method seems to work quite nicely 

s DeMaw and McCoy "Learning to Work with 
Semiconductors,” QST for April through 
November, 1974. DeMaw and Rusgrove, 
“Learning to Work with Semiconductors,” 
QST for April through November, 1975. 


up through the vhf spectrum. The hack- 
saw can even be used for some “casual” 
micro-strip uhf circuits for the 432-MHz 
band. No matter which method is 
chosen, keep the grounds short and 
clean, and many of the problems out- 
lined next will never occur! 

As an example of an rf circuit to 
debug, consider the rf power amplifier 
shown in Fig. 21. We’ll assume that a 
driving power from a VFO or mixer of 1 
mW is available, and that an ultimate 
power output of 2.5 watts is desired. 
Hence, a total gain of 34 dB is needed. 
While this gain could easily be obtained 
with only two stages, the use of a third 
stage will give us a much better chance 
of realizing unconditional stability. Two 
Class A stages are used to drive a Class C 
power amplifier. The base of the final 
amplifier is matched by means of an L 
network, and a single pi network is used 
for the output impedance trans- 
formation. 

The first step in testing such a design 
is to get a source of rf drive. Although 
die VFO which will eventually be used 
could serve to excite the amplifier, an 
equal approach would be to use an 
existing QRP transmitter. For example, 
one of the units from the preceding 
section would do the job, except that 
the power output is too high. This is 
easily remedied with a step attenuator 
of the kind outlined later on. The 
attenuator is adjusted for 1 mW of 
output, and we are ready to proceed. 

Only die first stage is attached to the 
signal source. The output link from LI 
is attached to a short length of coaxial 
cable which is run to a simple power 
meter. Power is applied to the first stage 
and Cl is tuned for maximum power 
output. Here is where some of the more 
subtle effects may rear their ugly heads. 
As Cl is tuned there should be a single 
well-defined peak, assuming the tuned 
circuit cannot be tuned to a harmonic 
of the input frequency. If the tuning is 
not smooth and well defined, the stage 
may be self-oscillating. The power out- 
put should disappear completely, of 
course, when the input drive is removed. 
At this time the stage should be checked 
for spurious output. The best amateur 
instrument for this is probably an ab- 
sorption wavemeter. Another useful 
tool is a bc-band receiver. If low- 
frequency oscillations are taking place, 
spurious responses may be heard while 
tuning from 550 to 1650 kHz. 

The Bandaids which may be applied 
to cure unwanted oscillations are many 
and varied. If spurious outputs (spurs) 
are noted in the low-frequency region or 
near the operating frequency, they may 
often be eliminated by placing a resistor 
in series with the base and/or the 
collector of the stage, typically 10 to 22 
dims. Also, reducing the stage gain may 
help a great deal. In this case the gain 


28 Chapter 2 



Fig. 21 - Circuit of a three-stage amplifier for use with text discussion of debugging. 


can be lowered easily by increasing the 
value of the 47-ohm emitter resistor. 
Varying the value of R1 should have 
little effect when the stage is being 
driven from our 50-ohm attenuator. 
However, it may add greatly to the 
stability when the VFO is tied into the 
system later. 

If vhf parasitics are observed with 
the wavemeter, they can be cured by 
means of the base or collector resistors 
mentioned above. Another solution is 
the use of a ferrite bead in either of 
these positions. If a clean layout is used, 
and proper bypassing is insured, vhf 
spurs are rarely a problem in hf trans- 
mitters. 

Since we are using three stages in 
this amplifier, and ultimately need only 
a gain of 34 dB. probably a good 
amount for the first stage would be 13 
dB. Hence, an emitter resistor which 
would yield an output of about 20 mW 
should be chosen. 

Once the first stage is operating 
properly, the second stage is built and 
connected. Since its output is meant to 
drive the base of the final stage, prob- 
ably the most effective way to test the 
system would be to build the final 
amplifier, but leave the output tran- 
sistor temporarily out of the circuit. 
With power applied to the first two 
stages of the amplifier, tire voltage is 
monitored across R3 with an rf probe 
and a VTVM. Typically, R3 will be 
approximately 39 to 56 ohms, or per- 
haps even less. C2 is tuned for maxi- 
mum power delivery to R3. The tuning 
of Cl is also checked. As before, tuning 
should be smooth. If spurs are observed 
the same Bandaids are applied to the 
second stage. The power delivered to R3 
should be around 200 mW. If this level 
is exceeded, the emitter resistor at Q2 
can be increased in value. Also, R2 is 
chosen to obtain the desired output 
from Q2 . 

When the first two stages are oper- 
ating properly, it will be time to add the 
final amplifier. Transistor Q3 is placed 


in the circuit, a 50-ohm power meter is 
used io terminate die rig, power and 
drive are applied, and the system is 
tuned. As before, all tuning is for 
maximum output. C2 will require re- 
tuning because the termination of tire 
second stage has changed with the ad- 
dition of the final -amplifier transistor. It 
may be desirable to increase the value of 
R3 in order to get more drive into the 
final amplifier. On the other hand, if 
there is the slightest sign of instability, 
the value of R3 should be reduced. 
Great care should be taken to ensure 
that the lead length of the emitter of 
the final stage is as short as possible. If 
the mounting method in a heat sink is 
such that a long lead is needed for Q3, 
make the connection with a relatively 
wide strap. A scrap of pc board or 
flashing copper can be used effectively 
for this. The 2N3553 used at Q3 has an 
fr of 400 MHz. If the emitter lead were 
as much as half an inch in length, vhf 
oscillations could almost be guaranteed. 
They would be observable with a wave- 
meter coupled near the final amplifier. 
However, they might not be observed at 
the output port due to the low-pass 
nature of the output network. 

If low-frequency oscillations are 
noted, they cannot be cured by adding 
the series resistance recommended for 
the first two stages, for such resistors 
would absorb too much power. The 
low-frequency spurs which might be 
occurring in the PA can be related to 
problems with the rf choke in the 
circuit. As suggested earlier, this choke 
should have a reactance (at most) of ten 
times the load resistance of the output 
stage. The electrolytic capacitor bypass- 
ing the supply to the last stage is then 
effective in killing the low-frequency 
spurs. If all else fails, a little resistance 
in parallel with the collector rf choke can 
be used to stop a low-frequency spur. 

Most likely the amplifier is operating 
nicely now. if the foregoing verbiage 
seems extensive, it is because of our 
attempt to cover ail bases. However, if 


careful construction practices are used 
(good grounding) and the gain-per-stage 
is kept down to a reasonable level, 
stability and smooth spur-free operation 
should be obtained without much 
trouble. When the board is mounted in 
the metal enclosure, and the transmitter 
is driven by the VFO (or whatever), it 
may be necessary to check the align- 
ment again, and ensure that stability has 
been retained. The pc board sitting on 
the bench may behave in a cleaner 
manner than the same board inside a 
metal enclosure. This is because energy 
may be radiated from the free board. 
However, when inside the metal box 
that radiated energy is reflected back 
into the box where it may interact with 
various parts of the circuit to cause 
unstable operation. 

One final test remains before the rig 
can be considered finished and ready for 
use. This is related to the output termi- 
nation used for testing. Typically, the 
load is a 50-ohm resistor of appropriate 
power dissipation, along with some 
means for rf-voltage detection. This 
load, if purely resistive, looks like 50 
ohms at all frequencies. Hence, the 
transmitter is terminated properly, not 
only at the operating frequency but at 
other frequencies. On the other hand, 
tlie typical antenna appears to be 50 
ohms (or thereabouts) at only one, or 
perhaps a few discrete frequencies. Else- 
where within the spectrum, it will be 
highly reactive. In some cases this can 
lead to instabilities, especially if emitter 
degeneration is used in the final stage. 

Testing for this condition is realized 
easily with a common ham-shack acces- 
sory a Transmatch or antenna tuner. 
Connect the transmitter to an absorp- 
tive type of bridge (see later chapter for 
details). The output of the bridge is fed 
to a Transmatch for the band in use, 
with file output of the Transmatch 
connected to the previously used 50- 
ohm wattmeter. The Transmatch is 
tuned for a balanced condition of the 
bridge. Then the bridge is removed from 
the system. An rf probe and VTVM are 
connected to the output of the trans- 
mitter and power is applied to the 
system. The rf voltage observed should 
be nearly identical to that observed with 
file broadband termination. When the 
various adjustments in the transmitter 
are tweaked, they should produce a 
smooth, stable variation in output, 
identical to that observed with the 
broadband termination. Any departures 
from these results are indicative of 
stability problems. Incidentally, if the 
power observed in the wattmeter is not 
close to that measured earlier, the 
Transmatch may need a bit of work. 

If the experimenter has both courage 
and a replacement for the output tran- 
sistor, there is another worthwhile ex- 
periment which can be done with the 


Basics of Transmitter Design 29 








Fig. 22 — Schematic diagram of the 6-meter QRP transmitter. Resistors are 1 /2-watt composition. Capacitors are disk ceramic unless otherwise 


noted. 

Cl, C2 — 30-pF trimmer capacitor. 

J1 - Two-circuit phona jack. 

J2, J3 — Phono jack or SO-239 fitting. 

J4 - Insulated jack for 12-volt input. 

LI - 10 turns No. 28 enam. wire on Amidon 
T-37-6 toroid core. 


L2 — 1 turn same wire over LI winding. 
L3 - 9 turns No. 28 enam. wire on T -37-6 
toroid core. 

L4 — 2 turns same wire over L3 winding. 
L5 — 6 turns No. 22 enam. wire on T-50-6 
toroid core. 


RFC1 - 15-hH choke. 

RFC2 - Two Amidon miniature ferrite beads 
on wire lead. 

Y1 — 50-MHz, third-overtone crystal (Inter- 
national Crystal Mfg. Co. type EX or 
equiv.). 


test setup outlined. The game is quite 
simple: Grab the controls on the Trans- 
match and twist them to grossly im- 
proper settings- That is, settings which 
would yield very high VSWR at the 
input to the Transmatch. If the output 
transistor survives this rather violent and 
potentially destructive test, the project 
is pretty well finished. It is then safe to 
use the transmitter in a fairly casual 
way, even with in-line type VSWR 
bridges for antenna adjustments. If the 
output stage does not survive, the blown 
transistor is replaced. The transmitter is 
still quite usable, but should be used 
only with something close to a proper 
termination. Furthermore, the rig 
should be used only with Transmatches 
which are tuned with an absorptive 
bridge. 

A 6-Meter QRP CW Transmitter 

When the universal QRP rigs de- 
scribed earlier were built, it was in- 
tended to include a 6-meter version 
along with the other designs. However, 
when construction was started, several 
problems occurred. The most severe one 
was that the 50-MHz crystal oscillator 
could not supply sufficient output to 
drive the final stage when it was biased 
to yield good stability. The next at- 
tempt was to try to combine two of the 
single-sided boards used for the rest ol 
the “universal” rigs. This also caused 
problems - the grounding was not good 
enough- Finally, it was decided to build 
a separate rig for 6 meters, apart from 
the designs for the lower bands, using 
double-sided board. The result is shown 
in Figs. 22 and 23. 

A three-stage circuit is used for the 


6-meter design. The crystal oscillator is 
a third-overtone circuit of the kind 
outlined earlier. The emitter resistor was 
increased from the usual 220 to 1000 
ohms in order to reduce the crystal 
current and improve the stability. The 
crystal oscillator is not keyed. 

Oscillator output is taken from a 
one-turn link and is applied to a keyed 
Class A buffer. This stage operates with 
fairly high gain due to the grounded 
emitter. Bias stability is achieved 
through the negative feedback at dc 
realized with the biasing scheme shown. 


The current is 15 to 20 mA. and the rf 
output from the buffer is about 50 mW. 

The final amplifier is a Class C 
2N3925. This device is specified for 
12-volt operation as an rf power ampli- 
fier in the 175-MHz region, and is 
capable of several watts of output. In 
this design, the power output was held 
down to a bit over 1 watt in order .to 
permit battery operation. The design of 
this stage was performed using the 
guidelines offered earlier, with the ex- 
ception that some additional decoupling 
was included in the form of a pair of 


Fig. 23 — Photograph of the vhf cw transmitter. The circuit board at the upper right con- 
tains the 1-watt 50-MHz transmitter of Fig. 22. The crystal oscillator is at the right end of the 
board and the output circuit is at the left. The stud-mount transistor is bolted to a small 
piece of circuit board, the latter of which is soldered to the main board. The remaining three 
pc boards form a similar design for the 2-meter band. The wafer switch accommodates T-R 
switching and band changing. 



30 Chapter 2 





ferrite beads on the collector supply 
line. A 2N3553 would probably serve 
nicely as a substitute for the outpui 
transistor used. 

The transmitter was enclosed in a 
small aluminum chassis box along with a 
switch for transmit-receive switching. 
Also included in the box is a crystal- 
controlled transmitter for 144 MHz. 


The design is similar to that described 
for 50 MHz. They can be seen in the 
photograph of Fig. 23. An alternative 
approach to packaging would be to 
include a simple crystal-controlled re- 
ceiving converter in the box with tire 
transmitter. 

Using only a 2 -element Yagi an- 
tenna, this transmitter has yielded 


several contacts over 1000 miles away. 
The reports were always compli- 
mentary. A frequent comment was that 
the rig provided “The cleanest cw signal 
ever heard on 6.” Perhaps this is not as 
much a testimonial for this transmitter 
as it is a commentary on the poor- 
quality cw signals often found on 6 
meters! 


Basics of Transmitter Design 


31 


Chapter 3 


More Transmitter Topics 


E^mphasis in this chapter will be on 
the more elaborate and practical con- 
siderations of transmitter design. We 
will treat VFOs, frequency multi- 
plication and mixing - all means of 
adding frequency coverage to a trans- 
mitter, beyond that which is reasonable 
for the crystal-controlled rigs in tire 
previous chapter. 

Several design examples are given. 
They are intended to illustrate the 
methods outlined in the text and are 
also suitable for duplication. Additional 
examples are given in later chapters. 

Building and Using VFOs 

In chapter 2 emphasis was placed on 
the use of crystal-controlled oscillators. 
The approach is ideal from a cost and 
circuit-simplicity outlook. However, 
there are occasions in operating where a 
VFO provides a necessary flexibility 
which is not possible with VXOs and 
simple crystal oscillators. A VFO per- 
mits greater effectiveness during low- 
power work, especially if crowded band 
conditions prevail. However, inclusion 
of a VFO compromises miniaturization 
and battery drain. Also, frequency sta- 
bility is more difficult to realize when a 
VFO is used in preference to a crystal 
oscillator notably when the equip- 
ment is designed for field use where the 
temperature environment may change 
markedly. It is of paramount impor- 
tance, therefore, to design for the best 
stability possible with ordinary circuits 
and components. 

VFO Design Philosophy 

As the radio amateur reviews the 
ham magazines, he finds a large number 
of VFO designs. The more extensive the 
search, the less rigid may be the con- 
clusions reached. Some of the popular 


circuits have names like Colpitts, Clapp, 
Seiler, Vackar and Hartley. Many of 
these designs are given in standard refer- 
ence books. 

VFO performance requirements are 
varied and many, and depend upon the 
intended application. For use in a 
typical transmitter the major need is 
that the oscillator have good long-term 
stability. By long term we mean that the 
oscillator should have a constant average 
frequency for periods of a second and 
longer. For critical receiver applications, 
and for most transmitters, the oscillator 
should have good short-term stability 
and low noise. In this chapter we have 


Fig. 1 - Block diagram of an LC oscillator. 


concerned ourselves mainly with the 
long-term stability matter — the 
“wanderies.” The problems of short- 
term stability, phase noise, and the 
“wobblies,” as well as a-m types of 
noise, are covered in the receiver chap- 
ters. 

Fig. 1 shows the block diagram of an 
oscillator. The basic components are a 
resonator (tuned circuit), an imped- 
ance-matching network, an amplifier 
and a second impedance-matching net- 
work. The two matching networks may 
include phase-reversing properties, de- 
pending on the nature of the amplifier. 
Typically, these networks are merely 



32 Chapter 3 



capacitors between the tuned circuit 
and the amplifying bipolar transistor or 
FF.T. The usual tuned circuit contains 
an inductor and capacitors, with the 
impedance-matching capacitors often 
being part of the resonator. Further- 
more. the parasitic capacitors of the 
transistors are, to some extent, part of 
the resonator. The better oscillators are 
those which use high-quality com- 
ponents throughout, such that changes 
in temperature do not change the fre- 
quency of tiie resonator. The sources of 
heat which can cause this drift include 
not only the external environment, but 
the heat created by the rf energy circu- 
lating in the loss elements qf the tuned 
circuit. 

There are a number of methods for 
matching into and out of the tuned 
circuit. The gentlemen who have studied 
the various methods now have their 
names attached to the configuration 
that they found most interesting. In 
general, the configuration chosen by the 
builder is secondary to considerations of 
component quality and fundamental 
design. 

The conditions for oscillation in a 
circuit of the type shown in Fig. 1 are 
described by the Barkhausen criterion. 
These conditions are related to Fig. iB 
where the feedback loop is opened at 
one point. Assume that the loop is 
opened at the input to the amplifier and 
that a signal is applied to the input of 
the amplifier. The conditions for oscil- 
lation (when the loop is closed later) 
are (1) The output signal after amplifi- 
cation and filtering should have an 
amplitude which is greater than the 
original signal and (2) the phase of this 
output signal should be exactly the 
same as that of the input signal. 

The first criterion specifies the gain 
needed in the amplifier. It’s just that 
amount required to overcome the losses 
in the resonator. The second criterion 
defines the frequency of oscillation. The 
oscillator operating frequency will be 
that at which the phase shift in the 
resonator is proper to fulfill the require- 
ment. 

These are general conditions. They 
have applied here to the design of 
VFOs. However, they may also be 
applied to crystal oscillators, or to audio 
oscillators which use RC networks. 
While we will not attempt such an 
analysis in this text, many of the guide- 
lines which follow result from a careful 
application of this theory, along with 
empirical observations. 


voltage-gain buffering may be used after 
the oscillator. In cases where additional 
driving energy is required, a simple Class 
A low-level amplifier can be included. 

The solid-state VFO offers a distinct 
advantage over a tube type of VFO 
reducing heating. The efficiency is 
better, and 60-Hz fm is not as likely to 
occur in a transistorized VFO. because 
there are no filaments to heat. Finally, 
miniaturization is greatly enhanced by 
employment of transistors as opposed 
to tubes in VFO circuits. 

It is beyond practicality to describe 
all of the VFO circuits which can 
provide good stability. Additional data 
not offered here can be obtained from 

i ii Rad,U Ama,eur ' s Handbook. We 
shall emphasize several circuits, all of 
which are easy to build and adjust. 

Long-term stability is attainable by 
adhering to some simple guidelines. 
Rule No. 1 is to use only that amount 
of feedback necessary to assure quick 
oscillator starting and minimum pulling 
by external load changes. Rule No. 2 is 
to bias the oscillator at a power level no 
greater than that needed for a specific 
output amount - generally, 10 mW or 
less of output power. Thi higher the dc 
input power to the oscillator, the 
©•eater the internal heating. Therefore, 
the rf currents flowing in the fre- 
quency-determining components ( L and 
C units) will be more pronounced. The 
higher the rf current flow, the greater 
the internal heating of capacitors and 
magnetic core materials. This leads to 
unwanted changes in operating fre- 
quency. So, in the present vernacular, 
keep it cooil 


Design Guidelines 

Some of the more common VFO 
circuits, such as the Colpitts and Clapp 
varieties, can be made stable enough for 
most amateur work, and the output 
levels will be ample for ordinary applica- 
tions. This is true even though unity- 


Components 

Temperature-stable capacitors 
should always be used in a VFO except 
where drift compensation is desired. 
Among the best low-cost capacitors 
available to amateurs are the dipped 
silver-mica and polystyrene varieties. 
The latter, generally speaking, have a 
much tighter tolerance to changes in 
temperature, and are highly recom- 
mended. Silver-mica capacitors are 
rather unpredictable with regard to 
temperature effects. Some may exhibit 
positive drift, while others from the 
same manufactured batch may change 
value in the opposite direction. Still 
others may be very stable in the pres- 
ence of changing temperature. This 
phenomenon has not been noted when 
using polystyrene capacitors in ARRL 
lab experiments. NPO ceramic capacitors 
are used in some VFO circuits, single or 
in combination with micas or poly 
units, with good results. 

I he VFO inductor should be ngid 
and of relatively high Q. Whenever 
possible, the coil should be without a 
magnetic core (iron or ferrite), as tem- 
perature changes will affect to some 


degree the permeability of the core 
material. Such changes will shift the 
inductance and, hence, the frequency. 
No matter what materials are used, the 
wire on the coil form should be cemented 
securely to the form by means of Q dope 
or some other high-dielectric compound. 
The inductor should not be mounted 
near any component that radiates heat. 

Toroidal inductors (magnetic core) 
are perhaps the most prone to changes 
in characteristics as the ambient tem- 
perature shifts. They should be used 
only in VFOs that will be operated in a 
■airly constant temperature environ- 
ment. The most stable toroid core ma- 
terial is the SF kind (Amidon type 6). 
Slug-tuned inductors are a better choice 
man toroids. They should be chosen 
and operated so that the slug barely 
alters the coil winding at resonance. 

I tie farther into the winding the slug is 
placed, the more pronounced the un- 
wanted temperature effects. 

The variable capacitor in a VFO 
should be mechanically stable, and 
should rotate smoothly with minimum 
torque applied. A double-bearing type 
of capacitor is recommended. Brass or 
iron capacitor plates are less subject to 
temperature effects than are aluminum 
plates. Air-dielectric trimmers are pre- 
ferred over those with ceramic or mica 
materials. 

If a bipolar transistor is used as the 
active element in a VFO, it should have 
an f T considerably higher than the VFO 
operating frequency, say, a 250-MHz f T 
for a 7-MHz VFO. This minimizes phase 
shift in the transistor. Furthermore, the 
small-signal beta should be 10 or greater 
to minimize the amount of feedback 
needed for reliable oscillation. When an 
FET or MOSFET is used in a VFO, it 
should also be a high-frequency device, 
and the transconductance should be 
2000 or higher. A 2N4416 or MPF102 
JFET is suitable for VFOs operating be- 
low 30Mliz. An RCA 40673 or 3 N2 00 is 
fine for VFOs which employ MOSFETs. 

Other Considerations 

Lead lengths in a VFO should be as 
short as possible. Excessive lead lengths 
become unwanted “parasitic" induc- 
tances. In circuits where very low values 
ot L are used, long connecting leads 
become a significant part of the tuned 
circuit and can degrade the Q. As a 
result, the VFO may not oscillate, or 
when the chassis is stressed the leads 
may move and cause shifts in the 
operating frequency. In some designs 
the circuit-board foils become part of 
the tuned-circuit inductance, so the 
layout should be planned for short, 
direct connections. 

Double-sided pc boards are not re- 
commended in VFOs ... at least not 
in the frequency -determining part of the 
circuit. The pc board, if double-sided, 

More Transmitter Topics 33 



Fig. 2 — Schematic diagram of two common VFO circuits. Reactance values are given for the 
critical components. 


provides numerous unwanted capaci- 
tances wherever die circuit foils are 
formed. The dielectric material of the 
pc board (phenolic or glass epoxy) is 
not especially stable with regard to 
changes in temperature and humidity, 
and drift can result from the double- 
sided board approach. Also, capacitors 
formed in that manner will be relatively 
low in Q , and this can lead to poor 
oscillator performance. 

Finally, the VFO should be con- 
tained in an enclosure to isolate it from 
stray rf which originates in odier parts 
of a receiver or transmitter. This also 
provides thermal isolation. Unwanted rf 
coupling can seriously affect VFO per- 
formance. It should be noted that VFOs 
can oscillate at some If, hf or vhf point 
odier than die desired one, while still 


performing at the chosen frequency. 
The amplifier following a VFO should 
be operated into a constant load imped- 
ance and the output examined by means 
of a high-frequency scope (if available). 
The waveform should be nearly a pure 
sine wave. Random oscillations above 
the VFO operating frequency will be 
superimposed on the fundamental wave- 
form. The measures prescribed earlier 
(ferrite beads, bypassing, addition of 
low-value resistors) for correcting in- 
stability are applicable in VFOs as well. 

The operating voltage for a VFO 
should be regulated and well filtered. In 
most amateur circuits a Zener-diode 
regulator will suffice. It is not uncom- 
mon to see regulation applied to the 
VFO anti its buffer stages. The practice 
is a good one to avoid load changes 


caused by voltage fluctuations, as they 
may pull the oscillator. Three-terminal 
1C voltage regulators are also well suited 
to this application. Some of the newer 
units are no larger than a plastic transis- 
tor. 

Examples which show two of the 
oscillators under discussion are given in 
Fig. 2. Approximations are given for the 
reactances of /, and C in significant 
areas of the circuit. These are ball-park 
values, and will enable the builder to 
scale either circuit to a selected tuning 
range in the hf or mf spectrum. At Fig. 
2 A, Cl can be the main tuning capac- 
itor, with C2 serving as a padder for 
calibrating the VFO to the dial readout. 
The absolute values of Cl and C2 will 
be dependent upon the size of coupling 
capacitor C c and both Cf b capacitors. It 
will be necessary to determine the 
combined series capacitance value of C c 
and both C,„ units, then add that value 
to Cl and C2 to find die tuning range of 
the oscillator. LI is a fixed-value com- 
ponent in this case. 

Generally speaking, die output 
capacitor, C a , should be as small in 
value as possible, consistent widi ade- 
quate output voltage to excite the fol- 
lowing stage (buffer or amplifier). The 
fixed-value capacitors just discussed 
should be polystyrene types for best 
frequency stability, but selected silver 
micas can be used if the builder is 
willing to soldcr-and-try until some 
stable ones are found. 

The circuit of Fig. 2B shows a Clapp 
VFO which is a series-tuned form of die 
Colpitts. It has been proved quite stable 
when used from 1.8 to as high as 10 
MHz. The advantage in using a series- 
tuned gate tank is that greater in- 
ductance is required than widi the 
parallel-tuned type of tank. This means 
diat stray inductances have less effect 
upon circuit performance - an advan- 
tage. At 7 MHz the circuit at A requires 
approximately 3 /uH for LI . Conversely, 
die circuit at B will have an L2 value of 
roughly 6 /all at 7 MHz. 

Capacitors C3 through C6. inclusive, 
are in parallel at the bottom of L2 in 
Fig. 2B. The advantage in using several 
capacitors instead of one or two is that 
the rf current is divided among them, 
which lessens the internal heating of any 
one capacitor. This greatly enhances 
stability. Similarly, die builder could 
use paralleled capacitors for the Cf* 
units for the same reason. 

If the Barkhausen criteria for oscil- 
lation outlined earlier arc examined, we 
see that they predict die signal in an 
oscillator will always be increasing. This 
is, of course, impossible. Something is 
required in any oscillator to limit the 
amplitude of oscillation. 

In die LET oscillator of Fig. 2B. the 
output of the circuit is stabilized by 
means of diode CR1 . The diode rectifies 


34 Chapter 3 





the rf signal from the tuned circuit and 
charges the capacitors to some dc value. 
This bias reduces the gain of the ampli- 
fier until the output voltage is sta- 
bilized. The oscillator would operate 
without this diode. However, the 
limiting bias would then be developed in 
the gate-source diode of the FL'T. This 
not only tends to create harmonics in 
the output, but loads the tuned circuit. 
Further, since the source of the FET is 
not tied to ground, the oscillator will 
operate at higher amplitudes. The larger 
circulating currents in the tuned circuit 
will degrade stability. 

With both circuits of Fig. 2 it is wise 
to apply the least amount of operating 
voltage practical. That is, use no more 
regulated voltage than is necessary to 
assure reliable operation and adequate rf 
output. The lower the voltage the better 
the stability, generally speaking. When 
FETs are used, the supply should ex- 
ceed the pinch-off voltage of the device. 
A good voltage range is from 6 to 9, 
regulated. The tuned-circuit com- 
ponents should be housed in a shield 
enclosure, as shown by the dashed lines. 
It is good practice to enclose the entire 
oscillator circuit in a metal compart- 
ment when space permits. 

Practical examples of VFO circuits 
are presented later in this chapter. In- 
formation concerning the design of buf- 
fer stages was provided in chapter 2. 

Any of the circuits- shown may be 
tuned with varactor diodes instead of 
the more common mechanically variable 
capacitor. There are, however, some 
problems which may occur. First, the 
diode should always be biased in such a 
way that the rf voltage does not cause 
the diode to conduct. The simplest way 
to realize this is to utilize two varactor 
diodes in a back-to-back arrangement, as 
shown in Fig. 3. While this arrangement 
decreases the net capacitance of the 
diodes by one-half, it prevents signifi- 
cant current from flowing in them. The 
second precaution that should be taken 
is to ensure that the variable biasing 
voltage is as clean and stable as possible. 
Any drift or noise on this controlling 
voltage will show up as instability or fm 
noise on the oscillator frequency. 

Some Other VFO Circuits 

Shown in Fig. 4 and the photograph 
is an adaptation of a Sieler-type oscil- 
lator developed by W2YM ( QST for 
Dec., 1966). While silver-mica capacitors 
are shown in the circuit, we later re- 
placed them with polystyrene units, 
resulting in an improvement in stability. 
The constants given are for 3.5-MHz 
operation. 

While a MOSFET was used in the 
original W2YM circuit, this oscillator 
also functions well with a JFET. It may 
be scaled to a number of other fre- 
quencies. The constants for several 





Here is the simple 80-meter VFO. The T-68-2 toroid inductor is seen at the upper right, and 
the JFET oscillator is at the top center. At the lower left is a two-stage buffer amplifier with 
feedback. The air trimmer is switched into the circuit bv means of a diode, providing a fre- 
quency offset function when desired. 


More Transmitter Topics 35 







Layout of the 160-meter transmitter with VFO. The top circuit board contains the entire 
transmitter. The VFO section is at the left. Seen at the bottom of the photograph is a 
crystal-controlled 160-meter converter with a /-MHz i-f. Front panel controls are for VFO 
tuning, VFO spotting, and T-R control. A receiver antenna trimmer is also on the front 
panel. The remaining circuitry is for a solid-state power amplifier and T-R relay. 



Fig. 6 - Schematic diagram of the 80-meter JFET VFO. Cl is the main-tuning capacitor, the 
value of which is selected for the desired tuning range. C2 is adjusted for the desired offset 
amount, and is an air-dielectric trimmer. LI is a T 68 2 toroid core wound with 30 turns of 
No. 22 enamel wire. 


other frequencies are shown in Fig. 5. 

When miniaturization is more signifi- 
cant than extreme long-term stability, 
toroid inductors can be used. Shown in 
Fig. 6 is an 80-meter VFO which was 
developed for use in a compact portable 
transceiver (described later in the book). 
A JFET has been used in the W2YM 
circuit. An additional feature of this 
design is the inclusion of a diode switch 
to shift the frequency slightly. When the 
diode has no external bias applied at 
point A, the small variable capacitor, 
C2, will charge to a dc voltage such that 
virtually no current flows in the capaci- 
tor. However, when +12 volts are ap- 
plied to point A, rf current will flow in 
C2, making it part of the resonant 
circuit. A decrease of up to 2 or 3 kHz 
can be realized, depending upon the 
setting of C2. 

Shown in Fig. 7 is a simple Hartley 
oscillator. This circuit is of significance 
for two reasons. First, it is easily scaled 
to just about any frequency in the hf 
spectrum or lower. Second, it demon- 
strates that component quality and 
proper application of design funda- 
mentals are more significant than a 
detailed oscillator configuration. 

This oscillator was first bread- 
boarded using a large piece of Mini- 
ductor coil stock and a 200-pF double- 
bearing air capacitor, tuned to reso- 
nance at 3.5 MHz. The small 1-10 pF 
capacitor was adjusted for easy starting, 
but was replaced later with a 5-pF 
ceramic NPO unit. Even though the 
oscillator was tested on the open work- 
bench with no shielding, in a room 
where the temperature was changing 
rapidly, the maximum drift observed 
over a two-hour period was 50 Hz. The 
air capacitor was then replaced partially 
with a fixed-value silver-mica unit, re- 
sulting in degraded stability. A similar 
degradation was observed when the air- 
core inductor was replaced with one 
wound on a T-68-2 toroid core. Good 
stability was maintained, however, when 
most of the capacitance was replaced 
with paralleled 47-pF NPO ceramic 


♦ 6V 

REGULATED 



Fig. 7 - W7ZOI high-stability Hartley VFO 
circuit. 


36 Chapter 3 






Fig, 8 - Schematic diagram of the 160-meter VFO. Capacitors of fixed value are disk ceramic unless otherwise indicated. Resistors are 1/2- 
watt composition. Numbered components not appearing in parts list are numbered for pc-board layout purposes only. Rms voltaaes were 
measured a VTVM and diode probe. 


Cl - 35-pF air variable (Millen 28035MKBB 
or equivalent). 

C18. C19 — .001-jiF feedthrough capacitor. 
CR1 — Small-signal high-speed silicon diode, 
1N914 or equivalent. 

LI - Slug-tuned high-Q inductor, 25 to 58 


nH (Miller 43A475CBI, Q u - 180 at 
2.5 MHz). 

L2 — Slug-tuned, pc-board-mount inductor, 
10 to 18.7 *iH (Miller 23A155RPC or 
equivalent). 

Q1. Q2- Motorola JFET. 


RFC1, RFC2 - Miniature 1-mH rf choke 
(Millen J301-1000or equiv.). 

RFC3 — Miniature 2.5-mH rf choke (Millen 
J302-2500 or equiv.). 

VR1 - 8.6-V, 1-W Zener diode. 


units. The tap on the coil was 1 /4 of the 
way up from the grounded end. 

A Practical High-Stability VFO 

The circuit of Fig. 8 is patterned 
after the VFO used in a W1CER 10-watt 
cw transmitter for 160 meters which 
was described in QST for November of 
1974. Stability is such that in this 
model the drift could not be measured 
with ordinary laboratory-style fre- 
quency counters during tests in a rela- 
tively constant temperature environ- 
ment (68 to 78 degrees F). From a cold 
start (no dc applied) to an “on” condi- 
tion exceeding two hours, the frequency 
remained constant at plus or minus one 
Hz. The operating voltage was keyed 
while monitoring the cw signal from the 
VFO, and a chirpless note characteristic 
was observed. While the builder may not 
be able to duplicate this stability, the 
circuit should still yield much better 
than typical performance. 

With the LC constants shown the 
VFO tunes linearly from 1.8 to 1.9 
MHz. An imported vernier mechanism 
with a 0-to-100 dial scale provided 
1-kHz readout increments. Increased 
frequency coverage can be had by em- 
ploying a main-tuning capacitor which 
has a greater maximum capacitance 
amount. 

A Clapp circuit is used to permit a 
greater amount of inductance at LI 
than would be possible with a parallel- 
tuned gate tank. The advantages of this 
were covered in the VFO philosophy 
section of this chapter. To enhance 



MOUNT R12 ON 
FOIL SIDE OF 
BOARD ACROSS 
L2 PINS 


TO C18 ON BOX WALL 


TQ 09 ON BOX WALL 


Fig. 9 — Scale layout of the VFO circuit board 


More Transmitter Topics 37 











Fig, 10 - Schematic diagram of the 160-meter QRP transmitter. Capacitors are disk ceramic unless otherwise noted. Cl is an 80-pF air 
variable (main tuning). LI is a T-68-2 toroid core with 45 turns of No. 26 enamel wire. L2 and L3 are Amidon T-50-2 toroid cores wound 
with 23 turns of No. 26 emanel wire. RFC1 must be able to pass 0.5A of dc current. T1 is an Amidon FT-37-61 ferrite toroid = 125) 
with 25 primary turns of No. 26 enamel wire. The secondary contains 4 turns of No. 26 wire. Resistors are 1 /2-watt composition. 



The pi-network output tank is a 
simple low-pass filter which attenuates 
harmonic energy. The broadbanding re- 
sistor, R12, does not significantly de- 
grade the filtering action of the tuned 
circuit. Measurements showed that the 
second harmonic was down some 38 dB 
from the fundamental output, and the 
third harmonic was down in excess of 
45 dB. 

The VFO is enclosed in an rf-tight 
box made of double-clad pc-board 
material. C18 and C19 are feedthrough 
capacitors which are installed on the 
box wall. C19 is part of the output 
capacitance of the pi network. A pc- 
board layout is provided in Fig. 9. 

Although' the VFO is designed for 


160-meter use, it can be used in com- 
bination with a frequency-multiplier 
stage for 3.5-MHz operation. Al- 
ternatively, it can be modified for 
higher operating frequencies by taking 
the reactances of the various compo- 
nents and calculating new L and C 
values (see Fig. 2). The pc-board pattern 
is suitable for other operating fre- 
quencies. 

A 1-Watt 160-Meter Transmitter 
with VFO 

There has been a rebirth of interest 
in the 160-meter band. While the 
number of QRP enthusiasts on 160 is 
small, the band offers excitement and 
challenge to the low-power enthusiast. 


Many of the regular operators on “top 
band" are accustomed to receiving weak 
signals. Hence, they are able to dig into 
the noise for a contact. 

Shown in Fig. 10 is the circuit for a 
simple VFO-controlled rig for 160 
meters. The design is straightforward 
and illustrates many of the circuits 
discussed so far. The VFO is adapted 
from the one shown in Fig. 4. The VFO 
is followed by a feedback amplifier with 
a closed-loop gain of unity. This drives a 
Class A keyed buffer amplifier. This 
stage differs slightly from those dis- 
cussed earlier because a broadband, un- 
tuned output transformer is used. This 
output transformer is much like a tuned 
toroid, except that the unit is wound on 


38 Chapter 3 












Cl - Small 78-pF air variable. (Miller No. 
2109 dual-gang miniature with only 78-pF 
section connected was used here.) 

C3-C5. incl. — ,001-pF feedthrough type. 

C6 — 100-pF mica compression trimmer. 

CR1 - Silicon switching diode, 1N91 4 or 
equivalent. 

J1-J4, incl. — Panel-mount coaxial jacks of 
builder's choice. 

K1 - Two-pole, double-throw. 12-volt, low- 
current relay. (24-V P&B KHP17D12 used 
here, with spring tension reduced for fast 
pull-in at 12 V.) 

LI - Slug-tuned coil with O u of 80 or more, 

6 pH nominal. (Miller 42A686CBI used 
here.) 

L2 - Pc -board-mount slug-tuned coil, 3.2 pH 
nominal. (Miller 23A476RPC used here.) 

J. W. Miller Co., P.O. Box 5825. Compton. 
CA 90224. 

L3 — 1 7 turns No. 26 enam. wire to occupy 


a ferrite core. Most of the toroids used 
by builders of solid-state gear are of 
powdered iron and have a relative per- 
meability of around 10 or less. The 
ferrite core used here (available from 
Amidon Associates) has a permeability 


total area ot Amidon T-50-6 toroid core 
(13 «H). 

L4 — 21 turns No 26 enam wire to occupy 
total area of T-50-6 toroid core, tap at 
6 turns from collector end. 

L5 — 12 turns No. 26 enam wire to occupy 
total area of T50-6 toroid core 

L6 — 1 1 turns No. 20 enam. wire to occupy 
total area of T-68-2 toroid core (0.9 pH). 

L7 — 1 3 turns No. 20 enam. wire to occupy 
total area of T-68-2 toroid core (1.2 pH). 

L8 — 8 turns No. 20 enam. wire to occupy 
total area of T-68-6 toroid core (0.5 pH). 

L9 — 10 turns No. 20 enam. wire to occupy 
total area of T-68-6 toroid core (0.55 pH). 

LI 0 — 25 turns No. 26 enam. wire to occupy 
total area of T-50-6 toroid core (2.4 pH). 

Q1, Q2, Q8 - Motorola transistor. 

Q3, Q4. Q9. Q10 — Surplus 2N2222 or equiv- 
alent. 

Q5, Q6, Q7 — RCA transistor. 


of 125. The reason that high per- 
meability is desirable for a broadband 
design is that high inductance may be 
realized with a relatively small number 
of turns. With a small number of turns 
the capacitance between turns is low 


R2 — 1000-ohm linear-taper control. 
RFC1-RFC4, incl. — Miniature rf choke 
(Millen J301 series or equivalent). 
RFC5-RFC10, incl. — 40-pH low-Q rf choke. 
Five turns No. 26 enam. wire on Amidon 
jumbo ferrite bead. 

51 - Subminiature slide switch. SI A and SIB 
each spdt. SIC and SI D single dpdt unit. 
(Radio Shack switches. See text.) 

52 - Spst miniature toggle switch (Radio 
Shack). 

53 - Dpdt miniature toggle switch (Radio 
Shack). 

T1 - Broadband 1:4 toroidal transformer. 

Ten bifilar-wound turns No. 24 enam. wire. 
8 twists per inch, to occupy entire area of 
two Amidon FT-61-301 ferrite toroid 
cores (stacked one atop the other). 

Amidon Associates, 12033 Otsego St., 

N. Hollywood, CA 91607. 


enough that self-resonances are avoided. 
Broadband performance is enhanced 
further by the fact that ferrites exhibit a 
permeability which is a decreasing func- 
tion of frequency. The transformer is a 
conventional type in contrast to the 


More Transmitter Topics 39 






Fig. 13 - Close-up view of the 20/40-metor 
solid-state transmitter. The cabinet is home- 
made from 1/1 6-inch aluminum stock. A 
cover was made from perforated aluminum 
which was obtained at a flea market. Kurz- 
Kasch aluminum knobs are used on the con- 
trols. The large knob on the vernier drive was 
cut down on a lathe to make it thinner, and 
to permit the set screw to mate with the drive 
shaft. An SWR-indicator meter is seen at the 
upper right. Green tape labels identify the 
controls on a green panel. 


transmission-line types described later in 
this book. 

The Class C output amplifier differs 
from those described earlier. First, the 
GE D-44C6 used for the final stage has 
an F T of over 50 MHz: The available 
gain is high. This could lead to in- 
stabilities. Stability was obtained by the 
addition of a small value of capacitance 
across the base-emitter junction. The 
second departure from the norm was in 
the design of the output network. We 
are ahead of ourselves a little here, for 
such designs have yet to be described. 
However, in this case we used what 
appears to be a typical half-wave filter. 
This is merely a double pi network, each 
section having a Q of 1. Usually it is 
designed for a termination of 50 ohms. 
In this case an impedance of 50 ohms is 
then presented to the collector. The 
unusual aspect of the network shown is 
that it was designed for a termination of 
35 ohms. This was done so that a 
number of available 5000-pF silver-mica 
capacitors could be utilized. We then 
take advantage of the characteristic of 
the half-wave filter wherein it behaves 
like a half wavelength of transmission 
line. The result is that a 50-ohm ter- 
mination on the output yields a 50-ohm 
load which is presented to the collector. 
More data will be presented later about 
the design of these networks. 

The output of this transmitter is 
approximately 1.2 watts into a 50-ohm 
load, and it is flat across the entire 
160-meter band. However, it should be 
operated through a Transmatch so the 
rig will always see something close to a 
50-ohm termination. 

A 3 X 6 X 13-inch chassis was used 
to house the transmitter, a crystal- 
controlled converter, an rf power am- 
plifier with an output of 6 watts and 


suitable T-R switching. The receiving 
converter will be described in chapter 4. 

This package is similar to the unit 
described earlier for the 6-meter band: 
All of the required circuits are con- 
tained in one box (see photograph). The 
items needed to complete the station 
are a receiver in the hf range, a power 
supply, a Transmatch and keyer. This 
station design has worked well for bands 
which are operated on a sporadic basis. 
One-hundred sixty meters is used only 
during the winter months, but 6 meters 
finds heavy use during the late spring 
and summer months. A similar unit for 
2 -me ter cw is used in the ARRL June 
and September vhf QSO parties. 

A 20- and 40-Meter CW 
Transmitter with VFO 

Fig. 1 1 shows the VFO used in our 
10-watt two-band transmitter. It is pat- 
terned after the 160-meter VFO of Fig. 
8. Only the L and C values have been 
changed to increase the operating fre- 
quency. A different pc-board pattern is 
used, but only to enhance miniaturi- 
zation. C2, CR1 , RFC1.C3 and R1 have 


been included to offset the VFO during 
receive periods. In that manner the VFO 
can be kept operational during standby 
to assure stability (avoiding warm-up 
drift). Measured drift with this model 
(at 7 MHz) was 25 llz over an ambient 
temperature range of 68 to 75 degrees 
F. Stabilization occurred in 30 seconds 
after turn on. 

The offset circuit is actuated by 
application of 13-volts dc during stand- 
by. CR1 acts as a switch when sat- 
urated, placing C2 in parallel with Cl. 
The amount of frequency shift can be 
set by selecting a suitable value for C2. 

This design was described originally 
by W1CER in QST for March, 1975. A 
low-power Bruene-style SWR bridge has 
been added in the cabinet for utility 
when afield. The circuit was described 
in QST for April, 1959. Also, R1 was 
changed from 10,000 ohms to the value 
shown in Fig. 1 1 . The lower resistance 
value cured a slight chirp which oc- 
curred during the first cw character 
when the break-in delay circuit was 
actuated. 

Fig. 12 contains the circuit diagram 


Fig. 1 4 — Interior view of transmitter. The VFO box is at the upper right with its aluminum 
cover removed. Directly below the VFO is the sidetono module. The large assembly occupying 
the center of the chassis is the rf power strip. Three miniature slide switches are ganged by 
means of a pc-board strip (left on power module). At the upper left is seen the break-in delay 
assembly. Below it is the SWR-indicator module. 


40 Chapter 3 



Fig. 15 — Diagram of a bipolar-transistor fre- 
quency multiplier. 


of the main section of the transmitter, 

E lus peripheral items. The break-in de- 
iy and side-tone circuits can be elim- 
inated if manual switching is desired, 
and if side tone is not needed. The 
functions of K1 can be effected by 
means of a two-pole double-throw 
switch. 

A power output of 7 watts is avail- 
able from this circuit, indicating a PA 
efficiency (Class C) of 70 percent. This 
power plateau is ample for most field 
work. During a two-week DXpedition 
(ZF1ST) this transmitter was used to 
work the world on 20 and 40 meters. 
Simple dipoles were erected near the sea- 
shore on Grand Cayman Island, neither 
of which was more than 25 feet above 
ground. Power consumption at 13 volts 
is just under 2 amperes. 

The PA tank circuit consists of two 
double-section pi networks, fixed-tuned, 
and serving as half-wave filter-matching 
networks. Because these are low-pass 
filters, a slight amount of 7-MHz energy 
appears at the transmitter output during 
20-meter operation. Therefore, a 40- 
meter trap is used (L10) to provide 
clean output at 14 MHz. Drive control 
R2 was included to permit very low- 
power experiments (QRPp), and to re- 
duce transmitter output when driving 
external high-power amplifiers. Band 
changing is made possible by ganging 
three miniature slide switches which are 
mounted on the amplifier-compartment 
wall and operated by means of a strip of 
pc board which is coupled to a knob on 
the front panel (push-pull action). 
Photographs of the interior and exterior 
of tiie equipment are shown in Figs. 13 
and 14. With the VFO LC values given, 
the tuning range is 7 to 7.070 and 14 to 
14.140 MHz. Increased range can be 
obtained by making Cl larger in capac- 
itance. 

Frequency Multipliers 

The designs offered in tire preceding 


pages have utilized oscillators which 
operate at the same frequency as the 
output of the transmitter (Fig. 12 ex- 
cepted). Certainly for the usual crystal- 
controlled rig, this presents no prob- 
lems. However, for work in the amateur 
bands above 7 MHz it is better practice 
to operate the VFO at a lower fre- 
quency. The output of the oscillator is 
applied to a stage which multiplies the 
frequency of the input driving signal. 
The major advantage of such a scheme is 
that die frequency multiplier provides 
excellent buffering. Stray rf from die 
final amplifier of a small transmitter has 
minimal effect if it is coupled into an 
oscillator operating at a different fre- 
quency. Of equal significance is that die 
builder can take full advantage of the 
harmonic relationship between the 
lower amateur bands and can build 
multi-band transmitters with relative 
ease. 

Most of die active devices used in 
electronics are linear in nature, at least 
for small signals. Mathematical analysis 
will show that the output of a linear 
amplifier contains only those fre- 
quencies present at the input, and 
nodiing more. Other frequencies, such 
as the harmonics we consider here, arise 
only from departures in linearity. 

Most writers stale that optimum 
performance is obtained from a mul- 
tiplier when it is biased and driven in a 
way diat the distortion products are 
maximized. However, the discussion 
usually ends there. The reason for this 
lack of data is really fairly obvious when 
one considers die measurements needed. 
The equipment required to evaluate a 
frequency multiplier is elaborate and 
expensive. Only in recent years has this 
gear become commonly available in 
even die better equipped electronics 
labs. 

In an attempt to fill this gap, a 
number of experiments were performed 
using state-of-the-art instrumentation. 
The basic unit was a Tektronix 7 LI 3 
spectrum analyzer in a model 7704 
oscilloscope mainframe. Even though 
sophisticated measurement gear was 
used to obtain die data which follows, 
the results are applicable to the amateur 
experimenter with his limited 
measurement capability. 

The first experiment was to evaluate 
a frequency multiplier of the type 
found in many published designs. Fig. 
15. A garden-variety silicon transistor 
was biased for 7 mA of dc collector 
current with no rf drive. With high-value 
rf-drive signals, the current may increase 
to 15 or 20 mA. The multiplier output 
contained a powdered-iron toroid, res- 
onated at 20 MHz. The performance as 
an amplifier, frequency doubler or a 
tripler, could be evaluated by applying 
drive from a signal generator at 20, 10 
or 6.7 MHz, respectively. The generators 



INPUT POWER. dBm 


Fig. 16 - Input power versus output power in 
dBm for a bipolar-transistor frequency 
doubler. See text for explanation of the 
curves. 


used in the experiments had 50-ohm 
output impedances; hence, die stage 
showed no instability. The circuit pro- 
vided a gain of 24 dB when operated as 
an amplifier. 

Shown in Fig. 16 are the results 
obtained when the stage was operated as 
a frequency doubler. The curves show 
output power as a function of input 
power. The data form may not be 
famdiar to the amateur. The powers are 
plotted in dBm, die unit which is used 
for most rf measurements within the 
electronics industry. Power in dBm is 
power referenced to 1 mW, Hence, 0 dBm 
is 1 mW, -30 dBm is a microwatt and 
+20 dBm is a tenth of a watt. The other 
atypical part of die data is that the 
component powers at the various fre- 
quencies of interest are plotted in- 
dividually. This allows us to compare 



i 

-<ol - 1 L- 

-*0 -10 O *10 


INPUT POWER. dBm 


Fig. 1 7 — Input power versus output power for 
a bipolar -transistor frequency tripler. See 
text for data concerning the curves. 


More Transmitter Topics 41 






Fig. 18 — Circuit of an FET frequency mul- 
tiplier. 


the desired doubler output (N = 2) with 
the fundamental feedthrough (JV = 1) 
and with the third harmonic of the drive 
frequency (/V = 3). The input power is 
not that actually delivered to the stage, 
but the power available from the genera- 
tor. There is a difference between the 
two. 

The results are quite revealing. We 
see that the doubler (Fig. 16) can 
provide output powers of up to 50 mW 
(+17 dBm) with a gain of 7 dB. How- 
ever, the multiplier is not very clean. 
The best suppression of undesired com- 
ponents in the output is only 16 dB. 
This occurs at outputs below the max- 
imum obtainable, a less than desirable 
situation when sophisticated test equip- 
ment is not available forevaluation. The 
performance could be improved sub- 



Fig. 19 — Illustration of a diode frequency 
doubler. 


stantially by increasing the selectivity of 
the output tuned circuit. This is most 
easily realized by tapping the collector a 
few turns from the V cc end of the 
output tuned circuit. A double-tuned 
circuit at the output, if designed prop- 
erly, would lead to an acceptable 
doubler. 

Shown in Fig. 17 are the results 
obtained when the stage was operated as 
a tripler. Performance is even worse 
than that of the doubler. The best 
suppression of undesired outputs was 12 
dB. This circuit would provide mar- 
ginally acceptable performance only if a 
double-tuned output tank were used. 

The next experiment is outlined in 
Fig. 18, where a JFET was evaluated. 
The first FET tried is typical of those 
used by the amateur, a 2N4416 with a 
pinch-off of about 5 volts. The results 
were discouraging. At high drive levels, 
the maximum output obtained was only 
+4 dBm, with spurious output down 
only 12 dB when operated as a doubler. 
Surprisingly, the results as a tripler were 
slightly better. With a drive of 10 volts 
pk-pk the output was still +4 dBm and 
die worst spur, the feedthrough of the 
6.7 -MHz drive, was down 16 dB. 

The FET was changed to a 2N4302. 
This device has a relatively low trans- 
conductance and more significantly a 
pinch-off of only 1.5 volts. When 
operated as a doubler the output power 
was quite low, only +1 dBm. However, 
all spurs were over 18 dB below the 
desired output. This occurred, again, for 
a 10 volt pk-pk drive. The performance 
as a tripler was extremely poor, al- 
though the behavior as a X-4 and as a 
X-6 multiplier was reasonable. This 
high-order multiplication is not recom- 
mended unless high-quality test equip- 
ment is available for evaluation and 
alignment. 

In view of the foregoing, it is no 
surprise that some amateurs encounter 
problems in building and adjusting gear 
for the higher hf bands. Furthermore, 
the problems are not limited to home- 
made equipment! A prime area where 
problems arise is in a 2 -meter fm rig for 
which a signal of 6, 8 or 12 MHz must 
be multiplied many times to arrive in 
the proper part of the vhf spectrum. 
Those vhf rigs which use double-tuned 
circuits throughout the multiplier chain 
usually have spurious outputs which are 
at least 45 or 50 dB down. Others rarely 
fare as well! 

All is not lost. The preceding pes- 
simism was intended to encourage the 
experimenter to strive for good designs. 
The key to building clean multipliers is 
balanced circuitry: At least some of the 
undesired output frequencies should be 
cancelled. Shown in Fig. 19 is a simple 
two-diode frequency doubler which was 
evaluated. Also presented are the classic 
waveforms for this circuit. We are much 


more familiar with this configuration as 
a full-wave power-supply rectifier than 
in rf circuits, but the same basics apply. 
The output rf choke will short the dc 
part of the output signal, effectively 
moving the zero reference up in die 
lower curve from the position shown. 

The balanced diode doubler shown is 
not included merely as an example of 
the effect of balanced circuitry. Shown 
in Fig. 20 are the output powers vs. 
available drive power for this circuit. 
While the diode doubler has a loss of 7. 5 
dB or more, the fundamental feed- 
through is as much as 41 dB down! 
Note that there are no tuned circuits in 
this multiplier. The performance ap- 
peared to be essentially the same over 
an output range of 1 to 50 MHz. The 
input transformer consists of seven tri- 
filar turns of Noi 28 wire on a ferrite 
toroid, 0.375-inch OD, and a permeabil- 
ity of 125. The diodes are silicon 
switching types of the 1N914 or similar 
variety. If a smaller core and hot-carrier 
diodes are used, the circuit will perform 
well into the vhf range. 

This simple diode doubler is used in 
a direct-conversion transceiver described 
later. Although a couple of tuned cir- 
cuits are used in later stages for im- 
pedance matching, no attempt was 
made to achieve good selectivity in the 
transmitter. Still, the 80-meter com- 
ponent in the output was measured at 
52 dB below the desired 7-MHz signal. 

The use of balance to remove un- 
desired frequencies from the output of a 
multiplier can be extended to stages 
with reasonable power output capa- 
bility. Two examples are shown in 
Fig. 21. A push-push doubler is shown 
at A. It uses a pair of 2N3904 tran- 



Fig. 20 — Input versus output power for a 
broadband balanced diode frequency doubler. 
See text for data on the curves. 


42 Chapter 3 






Fig. 21 - A push-push doubler is shown at A. The circuit at B is a push-pull tripler. 


sistors, For simplicity, only a bifilar 
winding is used as the input trans- 
former. This is otherwise identical to 
the transformer used with the diodes. 
With 10 mW of drive at 10 MHz, the 
output is tuned to 20 MHz with a 
resonant circuit using a powdered-iron 
toroid. The measured output power was 
50 mW. The spur components at 10, 30 
and 40 MHz were, respectively, down 
50, 40 and 3 1 dB. The collector ef- 
ficiency was 42 percent. 

Also shown is a push-pull tripler 
(Fig. 2 IB). This is identical to the 
doubler except that a balanced output 
circuit is used, tuned to 21 MHz. The 
output power was 32 mW with 10 mW 
of drive. The spurs at 7, 14 and 28 MHz 
were suppressed by 30, 55 and 46 dB, 
respectively, and die efficiency was 26 
percent. 

If proper methods are used, diese 
balanced multipliers may be used into 
the lower uhf region. A small cw trans- 
mitter was built with a 54-MHz crystal 
oscillator and three cascaded push -push 
doublers. All interstage networks are 
single-tuned, and a low-2 double-tuned 
filter is used on the output to yield 20 
mW at 432 MHz with only one detect- 
able spur, 55 dB down. 

Several ICs lend themselves well to 
clean frequency multiplication. This is 
because of the excellent inherent 
matching between monolithic tran- 
sistors, and diis enhances the balance. 


ICs investigated include the Motorola 
MC1496G, the RCA CA3046 and RCA 
CA3028A. 

The MCI 496 is a double-balanced 
modulator which is quite useful for 
mixing applications. It is used as a 
doubler by injecting the fundamental 
drive signal to both input ports simul- 
taneously. Although the drive level is a 
little critical, 60 dB of fundamental 
attenuation was observed with a single- 
tuned output circuit. The MC1496 is 
covered in more detail as a mixer in a 
later section. 

The CA3046 is an array of five 
transistors. Hence, four of the tran- 
sistors may be used to form a pair of 
multipliers of the type described in Fig. 
21. Other array-type ICs are worthy of 
experimentation. 

The CA3028A is a general-purpose 
1C consisting of a differential pair of 


transistors with a single transistor 
serving as a current source for the 
differential pair. If the current source is 
biased into saturation, the differential 
pair will serve well as a low-power 
push-oush doubler. This is depicted in 
Fig. 22. 

In general, any of the balanced 
multipliers outlined may be used. They 
all offer performance which is signif- 
icantly better than usually realized with 
single-ended configurations. However, 
there are problems encountered with 
balanced multipliers which are some- 
times difficult to diagnose without the 
aid of sophisticated instrumentation. 
These are related to imperfections in 
balance. 

Improper balance will result from 
two major causes. First is the problem 
of device similarity. For example, the 
push-push doubler of Fig. 21 will not 
perform as desired if one of the tran- 
sistors has twice the current gain of the 
other. For this reason, it is best to use- 
matched devices whenever these circuits 
are chosen. This is best realized through 
the use of integrated circuits such as the 
C A3 04 6 transistor array or the 
CA3028A differential amplifier. Even if 
a perfect match is obtained between the 
two devices in a balanced multiplier, less 
than optimum suppression of the fun- 
damental drive frequency will result if 
there is an asymmetry in the driving 
waveform. For this reason, the pre- 
ceding stage driving the multiplier 
should be a tuned amplifier, or should 
be a fairly clean Class A amplifier. An 
alternate might be the use of a low-pass 
filter such as the unit described at the 
end of the next section. 

It is not imperative that an IC be 
used in a push-push doubler, respective 
to matched transistor characteristics. 
Fig. 23 illustrates how a pair of 
2N2222A transistors is connected in 
push-push style and driven by a JFET 
source follower. T1 is tuned to 7 MHz, 
providing push-pull drive to the doubler 
transistors. Some forward bias is used 
on the doubler bases to increase the 
stage gain, but when driven the 
2N2222As operate in the Class C mode 
- essential to doubler action. R1 and 
R2 are chosen in accordance with the 
driving voltage available. In this example 



Fig. 22 — Schematic illustration of a CA3028A push-push frequency doubler 


More Transmitter Topics 43 




VOM for forward resistance. However, 
with unlike silicon diodes in the circuit, 
the suppression of the fundamental 
drive was measured at better than 40 dB 
down. With matched diodes, the sup- 
pression was nearly 60 dB. A number of 
these stages could be cascaded to form a 
multiband transmitter, starting with a 
VFO at 1 60 or 80 meters. 

Although a ferrite toroid was used in 
the input transformer, this could be 
replaced by a bifilar link on a previous 
tuned circuit. The output tuned circuit 
is chosen for the band of interest, and 
the output turns ratio is about 10. 

Mixer Design 

Although transistors have been used, 
the transmitters described thus far have 
been rather classic in design. That is, we 
have started with an oscillator which 
was (crystal-controlled or variable in 
frequency) followed by an amplifier. In 
some cases there has been a frequency 
multiplier or two somewhere in the 
chain. 

Today we find another approach to 
transmitter design which is becoming 
predominant. This is depicted in Fig. 
25. Instead of working directly with an 
oscillator at the output frequency, or at 
some sub-multiple of it, two oscillators 
are heterodyned in a mixer. The output 
of the mixer is tuned to a frequency 
which is the sum or difference of the 
two input frequencies. 

There are a number of advantages to 
using a mixer. First, stability is often 
improved. The reason for this is that 
one of the oscillators may be a highly 
stable crystal-controlled unit, while the 
other is variable in frequency. The VFO 
in the system may often be operat.ed at 
a relatively low frequency. This will 
enhance its stability. Furthermore, this 

Fig. 24 - Schematic diagram of a diode doubler followed by a low-level amplifier. T1 is an oscillator can run continuously. Hence, 

FT-37-61 ferrite toroid containing 10 bifilar turns of small enamel wire. one has to worr y about warm-up drift 

only once per operating session rather 
than every time a transmission is 

the FET was driven by a VFO lrom bias will not be required to ensure started. Another asset of a heterodyne 

which the output was approximately 10 adequate output from the doubler. approach to transmitter design is that 

mW. For the above reasons, the diode functions of keying and modulation are 

Dynamic balance of the 2N2222As doubler described earlier has appeal, well isolated from the critical variable 

is effected by means of control R3.The Shown in Fig. 24 is a general-purpose frequency oscillator. Finally, the mixer 

output waveform (14 MHz) is observed frequency doubler. The previously de- allows the frequency of a transmitter to 

on a scope and T2 is adjusted to scribed diode circuit is followed here by be controlled from the same oscillator 

resonance. Then, R3 is set for best a tuned amplifier. With 5 to 10 mW of that is used to control a companion 

waveform purity at 14 MHz. Unless the driving power, this "gain block” will superheterodyne receiver, making full 

doubler transistors are widely different provide up to 20 mW of output. The transceive operation practical, 

in their electrical characteristics, the diodes should be matched by means of a In spite of the advantages listed for 
balancing control will provide the de- 
sired effect. In laboratory tests of the 



Fig. 23 - A push-push frequency doubler using discrete bipolar transistors, and driven by a 
tuned-source JFET follower. Power output is approximately 20 mW. T1 has an impedance 
ratio of 1:1, primary to total secondary. T2 is tuned to the doubler output frequency. 

R3 is adjusted for dynamic balance of the two bipolar transistors (see text). 



circuit (Fig. 23), the output waveform 
contained no visible evidence of the 
7-MHz component after R3 and T2 
were adjusted as described here. A 
Tektronix 453 scope (50 MHz) was used 
It sufficient driving power is avail- 
able - 50 mW or more - the center tap 
of the T1 secondary winding can be 
connected directly to ground. Forward 


Table 1 



M-0 

1 

2 

3 

4 

5 

N-0 

0 

9 

18 

27 

36 

45 

1 

5 

4/14 

13/23 

22/32 

31 /41 

40/50 

2 

10 

1/19 

8/23 

17/37 

26/46 

35/55 

3 

15 

6/24 

3/33 

12/42 

21/51 

30/60 

4 

20 

11/29 

2/38 

7/47 

16/56 

25/65 

5 

25 

16/34 

7/43 

2/52 

11/61 

20/70 


Most cases show two numbers, representing sum and difference frequencies. 


44 Chapter 3 




Fig. 26 - Block diagram of a heterodyne fre- 
quency generator. 


mixers in a transmitter lineup, there are 
problems which make the design less 
than trivial. In many ways, the problems 
are akin to those encountered in our 
study of frequency multipliers. The 
mixer and the circuits following it 
should be designed in such a way that 
only the desired frequency is dominant 
in the output. Generally, if we have two 
input frequencies, /, and/ 2 ,a mixer 
output will contain components at Nf i 
t MJ\ where N and M are integer 
numbers starting at zero! 

Let’s consider an example, one 
which is typical because it is based on 
the frequencies used in many 20-meter 
receivers. Assume that we have a VFO 
in the region of 5 MHz, and the crystal- 
controlled oscillator is at 9 MHz. Some 
of the possible output frequencies are 
shown in Table 1 . 

The list was stopped arbitrarily at N 
and M = 5. However, it goes on (and on 
and on). Clearly, the spurious response 
is potentially worse than was the case 
with frequency multipliers where the 
only possible output frequencies were 
of die form N X /. 

If we study the list, remembering 
that our desired output is the 1:1 
response at 14 MHz, we see that mere 
filtering is not ample. For example, we 
see a 3:0 response at 15 MHz, and a 
couple of different spurs at 16 MHz, as 
well as a 1:2 response at 13 MHz. In 
spite of this, clean spur-free mixers can 
be built. The key to the design is the 
same as we encountered in building 
frequency multipliers - balance. That 
is, circuits are chosen which cause 
some components to be canceled in the 
output. With most mixers we will con- 
sider. the fundamental driving fre- 
quencies and their odd harmonics are 
well suppressed in the output, some- 
times by as much as 60 dB. The 
additional spurs may be suppressed by 
filtering and a judicious choice of input 
frequencies. 

Shown in Fig. 26A is the circuit for 
a double-balanced mixer using an 
MC1496G. By double balance, we mean 
that information at both of the input 
ports is suppressed from appearing in 


the output. (Often, in the generation of 
single sideband by the phasing method, 
a pair of balanced modulators is used 
with a common output. This is not what 
is usually meant by “double balance.”) 
The internal workings of the MC1496 
are shown at Fig. 26B. One signal is 
injected differentially on the bases of a 
pair of common-current sources. Since 
emitter degeneration is used at this 
input, it is usually the best point for 
applying a low-level signal where it is 
desired to preserve linearity. For 
example, this would be the place to 
apply a low-level ssb signal if such a 
transmitter were being built. 

The collectors of tire two signal- 
carrying input stages are then routed 
through four switching transistors. The 
stronger local-oscillator signal is applied 
to the bases of these switching tran- 
sistors. Using the component values 
suggested in the Motorola applications 


literature, the maximum current that 
ever need be switched is around 1 mA. 
Hence, fairly small local oscillator in- 
jection voltages are required to achieve 
proper switching action. Usually, signals 
of the order of 100 to 300 mV (rms) 
will be sufficient. In cw transmitters, 
the lower level signal can be as much as 
100 mV. In linear applications, how- 
ever, the signal at pin 1 should be less 
than this by 10 or 20 dB. Often, in 
linear applications, better distortion 
characteristics will be obtained by 
biasing the IC to larger currents. This is 
realized by decreasing the 10-kf2 resistor 
that connects to pin 5. The standing 
current in the IC is essentially twice the 
current (lowing into pin . 5. The 
Motorola data state that the chip should 
not run with more than 10 mA. 

Shown in Fig. 27 is the internal 
circuitry (A) and a mixer application of 
the RCA CA3028A (B). Although 



Fig. 26 - An IC mixer is shown at A. T1 is a toroidal bit ilar transformer, tuned to the desired 
output frequency- The internal circuit of an MCI 496G is shown at B. courtesv of Motorola. 


More Transmitter Topics 45 

















Fig. 27 — Circuit of a CA3028A single-balanced mixer at A. T1 is the same as for the circuit of 
Fig. 26. At B is the internal circuit of the CA3028A, courtesy of RCA. 


simpler than the previous circuit, this 
configuration has the disadvantage that 
it is only a single balanced mixer. That 
is, signals applied on pin 2 of the IC are 
suppressed in the push-pull output. 
However, the push-pull drive applied 
between pins 1 and 5 is not suppressed 
in the output. 

Fig. 28 presents the internal cir- 
cuitry (A) and a suggested mixer circuit 
(B) for the Tl SN-76514. This chip is 
similar to the MCI 496 in its operation, 
although the role of the rf and LO ports 
is reversed. The SN-76514 should be an 
easier “pill" to apply than the MCI 496 
since all of the biasing resistors are 
contained on the chip: One pays for this 
convenience by reduced versatility. 

In the sample circuits presented for 
the MC1496G and SN-76514, the out- 
puts are taken differentially between 
two collector terminals. However, if a 
builder is willing to accept reduced 
conversion gain, and this is usually 
acceptable, output may be taken from 
only one collector. The balanced prop- 
erties of tire chip will be retained so 
long as proper collector bias is main- 
tained. Using this design philosophy, it 
would be convenient to build a two- 
band transmitter. The band switching 
would be simplified by attaching a 
band-pass filter for each band to the two 
output collector points. This is shown in 


Fig. 29. Appropriate band-pass filters 
may be selected from the “filter cata- 
log" presented in the appendix. 

One of the really classic approaches 
to mixer design is to use diodes as the 
mixing elements. Two examples of 
diode-type mixers are presented in Fig. 
30. Like the other examples presented, 
these mixers are balanced. The two- 
diode mixer is single-balanced while the 
diode-ring mixer is double-balanced. 
Diode mixers exhibit loss instead of the 
gain associated with the other mixers 
presented. Impedance matching is crit- 
ical in diode mixers, and some spur 
responses are not well suppressed. On 
the other hand, diode mixers come into 
their own in broadband applications and 
in situations where wide dynamic range 
is desired. Most mixers of this kind 
utilize hot-carrier diodes, such as die 
HP-2800. However, for the hf region 
silicon switching diodes are often satis- 
factory. They should be matched for 
similar foward resistance. 

FETs of the junction and the MOS 
types may be used in transmitting 
mixers. However, they are used ideally 
in balanced configurations. While the 
dual-gate MOSFET is popular as a re- 
ceiver mixer, it has tire problem that 
harmonics of the local oscillator, par- 
ticularly even-order ones, are easily 
created within the device. This can 


cause serious problems with spurious 
responses unless good balancing tech- 
niques are used and careful filtering is 
applied. Additonal information on 
mixers is presented in the receiver 
chapter. 

Frequency Synthesis 

When we hear the term “frequency 
synthesizer,” we may think of the 
techniques used for frequency control 
of 2-meter I'm equipment. Narrow-band 
vhf-fm is a mode of amateur communi- 
cations which requires great frequency 
accuracy and stability. Hence, it is ideal 
for synthesis techniques. However, fre- 
quency synthesis is by no means limited 
to 2-meter fm. It appears that such 
methods will become predominant as 
the major means of frequency control in 
all high-performance amateur equip- 
ment. 

In the general sense, frequency 
synthesis is any process which elec- 
tronically operates on one or more 
frequencies to produce other fre- 
quencies. The mixers and frequency 
multipliers we have discussed earlier are 
examples of simple forms of synthesis. 
There arc, however, other methods 
which can be applied. 

It would be folly to attempt a com- 
plete treatment of synthesizers. Such a 
discussion would take us well beyond 
the relatively empirical scope of this 
volume. Nonetheless, synthesis methods 
are becoming so popular that some 
explanation is required. We will confine 
our discussion to two types of synthe- 
sizers which are of interest to the 
experimentally inclined amateur. 

The major advantage of frequency 
synthesis is stability. If one begins with 
a highly stable. crystal oscillator as the 
reference frequency, the output of the 
synthesizer using this reference will have 
a stability which is dependent upon the 
characteristics of the quartz crystal 
rather than a less stable VFO. If the 
system is well designed, the stability will 
be quite good. One of the simplest 
synthesizers the amateur can build con- 
sists of nothing more than a pair of 
crystal-controlled oscillators and a 
mixer. Each oscillator contains a bank 
of switchable "rocks." The advantage of 
a scheme of this kind is that the 
stability of crystal control is retained 
while great frequency accuracy is ob- 
tained. An additional characteristic, 
which may or may not be an asset, is 
the digital nature of the “tuning." Such 
digital techniques are useful for portable 
equipment designed for cold-weather 
conditions. 

As an example of this type of 
synthesizer, consider the block diagram 
of Fig. 31. Here, the two crystal oscil- 
lators are operated at 20 and 27 MHz. 
Each oscillator has five crystals avail- 
able. The two reference frequencies are 


46 Chapter 3 







Fig 28 — Circuit for a doubly balanced SN-76514 mixer, at A. The circuit at B shows the internal work 
ings of the 1C, courtesy of Texas Inslruments. The SN-76514 mixer 1C has been reidentified as 
TL-442-CN by Texas Instruments. II may be procured under eilher part number 


applied to a mixer with an output at 7 
MHz. A low-pass filter at the output 
ensures that none of the higher-order 
spurs are present. With an investment in 
only 10 crystals, 25 discrete frequencies 
in the 40-meter band will be available. A 
module of this sort would not be 
expensive to build, for CB crystals could 
be used in the 27-Mllz oscillator. 

Although frequency synthesizers 
using banks of crystal-controlled oscil- 
lators are fairly common, they are not 
as practical as might be desired. This is 
because a large number of crystals are 
required if versatility is desired. The 
techniques used to avoid this deficiency 
are usually based upon the phase-locked 
loop (PLL). 

There are a number of circuits which 
will serve as the critical element in a 
PLL (the phase detector). A phase 


detector is a three-port circuit, much 
like a mixer. At two of the ports (the 
inputs) two signals at the same fre- 
quency are applied. At the third port, a 
dc voltage appears. This voltage is pro- 
portional to the phase difference be- 
tween the two input signals. 

A simple PLL is shown (Fig. 32) in 
block-diagram form. The system in- 
cludes the phase detector, a reference 
oscillator, a voltage-controlled oscillator 
(VCO) and a loop filter. The phase- 
detector operation was defined above, 
and the reference oscillator could be, as 
an example, a stable crystal-controlled 
oscillator. The VCO is merely a VFO 
with the usual mechanically tuned ca- 
pacitor replaced with a varactor diode. 
As the voltage on the varactor is 
changed, the effective width of the 
depletion region of the diode changes. 


causing the diode capacitance to change. 
The loop filter is essentially a low-pass 
filter which tends to remove any ac 
components from the output of the 
phase detector. 

How is this system used to control 
frequency? The key to understanding 
PLL operation, at least on a rudi- 
mentary basis, is to recall that fre- 
quency is merely the rate of change of 
phase. That is. the phase of a signal 
from a highly stable oscillator is a 
constantly changing parameter. Once 
during each cycle of oscillation tire 
phase returns to some ‘‘zero-degree” 
reference point. Recall that the phase 
detector is a circuit which compares the 
phase difference between two signals. If 
the outputs of our two oscillators (the 
reference and the VCO) are exactly at 
the same frequency, there will be some 
dc voltage at the detector output. This 
dc level is proportional to the constant 
phase difference, whatever it may be, 
between the two oscillators. 

Assume now that the VCO starts to 
drift a little with respect to the fre- 
quency of the reference. Say, for ex- 
ample, the VCO tends to move in 
frequency by 1 Hz from that of the 
reference. If the two frequencies were 
indeed different by 1 Hz, the phase 
difference would be continually 
changing. That is, it would be a I -Hz ac 
signal. However, in our PLL. this never 
happens. As soon as the phase starts to 
shift the resulting dc signal from the 
phase detector is amplified (and fil- 
tered) in the loop filter and then applied 
to the VCO. The change in the dc 
control voltage on the varactor diode of 
the VCO is just that required to bring 
the frequency of the VCO back to that 
of the reference oscillator. The control 
voltage may be different from that 
present before the VCO started to drift, 
but the frequencies will be the same. 

The simple PLL shown in Fig. 32 has 
one flaw which may not be apparent 
immediately. It will, however, become 
painfully clear when one attempts to 
build such a unit. Assume, for example, 
that the crystal reference is at 1 MHz 


Fig. 29 - Illustration of an 1C mixer with 
output on two frequencies. Pins not indi- 
cated on the SN-76514 are connected as 
shown in Fig. 28. 



More Transmitter Topics 47 









Fig. 30 — At A. a two-diode mixer. A four-diode mixer is shown at B. The transformers are 
trifilar-wound on ferrite toriod cores. 



and that the VCO is capable of tuning 
from 0.9 to 1.1 Mil/, with the available 
voltages. Most likely, what will happen 
when power is first applied to the 
circuit is that the VCO will start oscil- 
lating at one end of the control range or 
the other, and it will stay there. With a 
100-kHz difference in frequencies, there 
will be no dc control voltage emanating 
from the phase detector just the ac 
signal at 100 kHz. What we must do is 
to initially "perturb" the VCO until it is 
momentarily at the same frequency as 
the reference. Then a suitable dc phase- 
controlled signal will exist which will 
cause the PLL to “lock up” and control 
the VCO. This perturbation is usually 
realized by additional circuitry which 
will cause the VCO to sweep over its 
range prior to lockup. 

A simpler and more convenient ap- 
proach to this problem is to replace the 
phase detector with a phase-frequency 
detector. This circuit provides a dc 


output which is a function of frequency 
difference prior to lockup. This signal, 
in combination with the smoothing ef- 
fects of the loop filter, will in effect 
generate the required sweep voltage. 
Once the VCO is near the frequency of 
the reference, normal phase-detector 
operation commences. An example of 
such a detector is the MC4044. The 
detailed operation of this digital circuit 
is rather complicated, but is well out- 
lined in the Motorola literature. 

Shown in Fig. 33 is a simplified 
phase-frequency detector which is built 
from a pair of D flip-flops and a NAND 
gate. A D type of flip-flop is a fairly 
simple device in comparison to many of 
the digital circuits used extensively in 
modern electronics. Whenever the posi- 
tive edge of a pulse appears at the clock 
input, C, the logical state present at the 
D terminal is transferred to the output, 
Q. Q is merely the opposite logical state 
from that at the Q terminal at any 


instant. A logical zero applied at the 
reset input, R, will always return the Q 
output to a logical zero. 

In the circuit shown, an SN-7474 
dual-D FF is used, in conjunction with a 
single NAND gate from an SN-7400 
quad two-input NAND-gate package. 
The D terminals are always tied to a 
logical one. Hence, whenever a positive- 
going pulse appears at either clock 
input, that flip-flop is set into a high 
output state. Each of the two flip-flops 
is clocked by one of the two input 
frequencies. The NAND gate is wired 
such that both flip-flops are reset to 
zero whenever both Q1 and Q2 are 
simultaneously at a logical one. 

Several sets of possible waveforms 
are shown in Fig. 33. At B are two 
different input frequencies with /, 
higher than f 2 . The appropriate levels 
for Q1 and Q2 are also shown. Of 
significance is the high average level of 
Ql. When this is smoothed out in the 
loop filter, we will have a dc signal 
coming from Ql which tells us that /, is 
higher than f 2 ■ 

The curves at C are similar, except 
that here f 2 is higher than j\ . We see 
that die average value of Q2 is much 
higher than Ql . 

Fig. 33D and E depicts and f 2 
equal in frequency, but out of phase 
with each other. As shown in the curves, 
the outputs at Ql and Q2 will tell us 
what nature and magnitude of the phase 
difference is actually present. In the 
case where exact phase coincidence oc- 
curs, the outputs from Ql and Q2 will 
both be very short positive pulses. 

Fig. 34 shows how this phase- 
frequency detector is interfaced with 
the loop filter. Note that the outputs 
used from the detector are Ql and Q2. 
The circuit is meant to be generally 
descriptive of the operation and lacks 
many of the interfacing details neces- 
sary to provide stable operation. These 
details will depend upon the final sys- 
tem configuration. 

As we study the simple PLL of Fig. 
32, the first impression we get is that 
the system is redundant. That is, why 
would one use an oscillator at 1 MHz to 
control another? Why not take the 
output directly from the reference oscil- 
lator, dispensing with all of the other 
circuitry? While the present system is an 
illustration, simple loops of this kind are 



Fig. 32 - Illustration of a basic PLL circuit. 


48 Chapter 3 




Fig. 33 - Representative phase-frequency detector using an SN-7474 1C and 1/4 of an SN 7400 1C. See text for details of illustrations 
8 through E. 


of value in some advanced systems. For mentioned long-term stability (the long-term variations in the frequency 

example one could arrange the circuitry "wanderies") and short-term drift (the are not necessarily related. That is. one 

and choose a proper phase detector such “wobblies"). Long-term drift is an in- may fight for long periods of time to 

that the outputs from the two oscil- stability which usually has its origin in remove the wanderies from a VFO, only 

lators were 90 degrees out of phase. The thermal effects. Short-term wobblies, on to find that he has designed a highly 

two outputs could then be used for the other hand, originate from noise in stable noise source. Noise considerations 

generation of ssb by the phasing the oscillator. Random variations in the are of major significance in the design of 

method. output of the amplifying device used in any phase-locked loop. In many synthe- 

A much more significant application an oscillator will cause minor variations sizers used by amateurs, a PLL has been 

of a simple loop of this kind relates to in the phase (and hence, the frequency) used to achieve a degree of long-term 

the noise characteristics of an oscillator, of an oscillator. The net result is that stability at vhf which surpasses that 

When we think of an oscillator, we our oscillator seems to provide a dis- found on even the lower hf bands, but 

envision a device which has an output at Crete frequency which is modulated by creates a signal which is excessively 

one discrete frequency. Perhaps we ac- noise. In this case, the modulation noisy. Casual application of PLL tech- 

knowledge the existence of a few liar- appears as a variation in phase of the niques can be quite disastrous, 

monies, but take a simplistic view of the oscillator. This pm or fm - the distinc- On the other hand, a PLL can be 
typical oscillator. Usually, this is justi- tion between the two is essentially used to clean up residual phase noise in 

fied. However, if one attempts to build nonexistent causes sidebands next to an oscillator. The simple loop of Fig. 32 

equipment which approaches the state the "carrier.” These noise sidebands could be a good example. If the refer- 

of the art (whatever that means), the may be the ultimate limitation in the ence oscillator were quite stable (long 

noise characteristics of the oscillator design of a wide dynamic-range receiver, term) and noise free, essentially all of 

must also be considered. as one significant example. this cleanliness could be impressed upon 

In our earlier discussion of VFOs we Unfortunately, the short-term and the output of the VCO which might 

otherwise be much less than clean. 
However, only those noise sidebands on 
the VCO which are separated from the 
VCO carrier by a frequency difference 
less than the bandwidth of the loop 
filter will be suppressed by the PLL. 

Let's now consider a somewhat more 
complicated synthesizer based upon the 
PLL shown in Fig. 35. This unit is 
typical of many units which have been 
implemented for 2-meter fm use. We 
Fig. 34 - Simplified schematic diagram of a loop filter for use with a phase-frequency detector. have shifted OUT reference frequency 

down to 1 kHz. This is easily done by 
starting with a crystal-controlled oscil- 
lator at 1 MHz, then applying the 
resulting signal to a divide-by-1000 cir- 
cuit. Typically, this would consist of 
three SN-7490 decade dividers. Sim- 
larly , the output of the VCO is applied 
to a frequency divider. Let’s assume for 
the moment that the VCO operates in 
the 6-MHz region and that the divider is 
set up to divide by 6000. If the VCO 
were right at 6 MHz, we would have two 
1-kHz signals being applied to our 
phase-frequency detector. The phase- 
proportional detector output would 
now be filtered in the loop filter and 
applied to the VCO. The VCO would 
move to the exact frequency required to 
Fig. 35 - Block diagram of a divide-bv-N synthesizer. achieve lock, where both inputs to the 




More Transmitter Topics 49 




phase detector have a stable, well- 
defined phase difference. 

A system of this kind is made 
“tunable" over a band of discrete fre- 
quencies by replacing the VCO-driven 
frequency divider with one which is 
programmable. That is, from the front 
panel of our synthesizer we could set 


switches which would cause the divider 
to, for example, divide by 6132 instead 
of 6000, causing the VCO to lock up at 
6.132 MHz. By changing the division 
ratio we pick the desired output fre- 
quency. In some kinds of synthesizers 
the divider in the reference-frequency 
chain is also programmable. 


It is worthwhile to consider the opera- 
tion of the detector in more detail. The 
reference frequency in this case is I kHz. 
As a result, once every millisecond the 
digital phase detector is pulsed by the 
reference. The phase detector serves the 
function of telling us whether the similar 
pulse from the programmable divider ar- 



Fig. 36 — Shown here is the schematic diagram of the 15-meter transmitter. Fixed-value capacitors are disk ceramic unless specified otherwise. 
Fixed value resistors are 1/2-W composition unless noted otherwise. Numbered components not appearing in the parts list are identified for 
pc-board layout purposes only. 


C3 - 47-pF polystyrene. 

C4, C5 - 240-pF polystyrene. 

C6 - 4- to 53.5-pF variable (Millen 22050 or 
equiv.). 

Cl 8 - 100-pF electrolytic. 25 volts. 

C22. C28 — 2.7- to 30-pF variable (Elmenco) 
461 or equiv.). 

C24, C27, C30 - 10-pF tantalum or electro- 
lytic, 25 volts. 

C31 — 25- to 280-pF variable (Elmenco 464 
or equiv.). 

CR1.CR2- 1 N91 4 or equiv. 

J1 . J2 — Coaxial connector, type SO-239. 

J3 - Phone jack (Radio Shack 274-280 or 
equiv.). 

J4, J5 - Binding post. 

LI - 6.05- to 12.5-pH adjustable coil (Miller 
42A105CBI or equiv.). 


L2 - 1 7 turns No. 28 enam. wire on Amidon 
T-50-6 core. 

L3 — 10 turns No. 28 enam. wire, center 
tapped, wound over L2. 

L4 — 1 7 turns No. 28 enam. wire on an 
Amidon T -50-6 core. 

L5 - 5 turns No. 28 enam. wire wound over 
L4. 

L6 - 30 turns No. 28 enam. wire on an 

Amidon T-50-6 core. Tap 10 turns above 
C23 end. 

L7 - 4 turns No. 28 enam. wire wound over 
L6. 

L8 — 30 turns No. 28 enam. wire on an 
Amidon T-50-6 core. Tap 7 turns above 
C26 end. 

L9 — 3 turns No. 28 enam. wire wound over 
L8. 


L10 - 22 turns No. 28 enam. wire 

L1 1 - 29 turns No. 22 enam. wire on an 
Amidon T -68-6 core. 

Q1.Q2 - Motorola MPF1 02 JFET or equiv. 

Q3, Q4, Q5 - 2N2222 transistor. 

Q6 - RCA 40082 transistor. 

Q7 — RCA 40977 transistor. 

RFC1, RFC2, RFC3 — 500-pH rf choke 
(Millen J-302-500 or equiv.). 

RFC4 1 6 turns No. 28 enam. wire on an 
Amidon FT-50-61 core. 

RFC5 - 1 1 turns No. 22 enam. wire on an 
Amidon FT-50-61 core. 

RFC6 — 6 turns No. 22 enam. wire on an 
Amidon FT-50-61 core. 

51 — Dpdt miniature toggle switch. 

52 - Spst momentary-contact push-button 
switch. 

VR1 — Zener diode, 9.1 volt, 1 watt. 


50 Chapter 3 




Fig. 37 Interior view of the transmitter. The VFO is in the compartment at the top The rf 
power strip is the lower module. 


rived before or alter the reference pulse. 
The output signal is a short pulse of the 
right polarity to ultimately cause the VCO 
to shift as needed to assure phase coinci- 
dence. The average of these pulses is our 
dc level. The purpose of the loop filter is 
to remove, as much as possible, the pulse 
or ac variations in the signal applied to the 
VCO. However, in designing the loop 
filter, we now encounter problems. First, 
if we are going to effectively filter out a ser- 
ies of pulses occurring at a 1-kHz rate, we 
must use a low-pass filter with a bandwidth 
of well under I kHz. This, unfortunately 
means that it is difficult to change frequen- 
cies. When we switch the programmable 
divider to a new ratio, the VCO will 
"hunt" for a short period, being driven by 
the proper frequency difference signal from 
the phase- frequency detector. If the loop 
filter bandwidth is as narrow as I Hz, the 
loop may take over a second to settle at a 
new frequency. A compromise bandwidth 
is usually used. 

No matter how narrow the filter 
there will be some pulse or ac com- 
ponent which will be applied to the 
VCO. Hence, the VCO is being fre- 
quency modulated by our 1-kHz refer- 
ence. With a suitably narrow loop filter 
the resulting sidebands are fairly well 
suppressed. However, when the VCO 
output is used to drive a frequency 
multiplier chain, as would be the case 
with a 2-meter fm transmitter, the 
suppression of the residual reference 
sidebands deteriorates. In general, the 
residual sidebands will come up by 6 dB 
every time the frequency is doubled. 

Another problem arises when we are 
forced to use an exceptionally low loop 


order to achieve good suppression of the 
bandwidth. As mentioned earlier the 
inherent noise sidebands in an oscillator 
can be suppressed only at separations 
from the carrier which are less than the 
loop bandwidth. In the case described 
the PLL does essentially nothing to 
make the VCO output quieter. When 
the VCO is applied to a multiplier the 
amplitude of the noise sidebands will 
again grow, just as the reference fre- 
quency sidebands did. The design of the 
VCO in the example considered must be 
extremely well done if the ultimate 
result is to be tolerable. 

One final point should be made 
about the design of the loop filter. In 
reference sidebands at the VCO, one 
might be tempted to use a compliciated, 
multisection low-pass filter of the kind 
used for audio filtering in a direct- 
conversion receiver, except, of course, 
having a lower cutoff frequency. In 
general, this approach is not viable. The 
reason is that any filter will exhibit 
maximum phase shift in any region 
where the attenuation is changing 
rapidly with frequency. The ultimate 
result of this phase shift is that the 
entire PLL may oscillate. These oscil- 
lations are detected experimentally as 
an ac component on the “dc” signal 
being applied to the VCO. 

While it is hard to generalize, the 
better PLL designs are those which use 
the highest possible reference fre- 
quency. Furthermore, it is desirable to 
operate the VCO at the highest reason- 
able frequency. Finally, heterodyning 
the VCO output to a desired output 
frequency is recommended over fre- 


quency multiplication. This is because 
of the degradation of the noise side- 
bands inherent with multiplication. 

Frequency -synthesis techniques 
offer great promise for future amateur 
equipment. However, great care is re- 
quired in the design if high performance 
is desired. 

A Deluxe 15-Meter CW Transmitter 
with VFO 

This circuit was described originally 
in QST for January, 1976, by 
WA1LNQ. Power output is approxi- 
mately 6 watts across 50 ohms when 
using a 12-V dc supply (1.3 A), and 7 
watts of output can be had at 13 volts. 
Frequency coverage is from 21.0 to 
21.250 MHz with the constants speci- 
fied in Fig. 36. An interior view is given 
in Fig. 37, and the outside of the 
assembled unit is shown in Fig. 38. 

The series-tuned VFO is fashioned 
after the circuit of Fig. 8, and the 
push-push doubler follows the lines of 
the circuit in Fig. 23 of this chapter. 
Stability is excellent at 21 MHz (less 
than 70 Hz from cold start to stabili- 
zation, requiring approximately two 
minutes). 

A spectral analysis of the 21-MHzrf 
output (at the 6-W level) shows the 
second harmonic to be down 45 dB, and 
the third harmonic is 55 dB down from 
21 MHz. The cw note is free of clicks 
and chirp. 

The VFO offset circuit (C2 and 
CR1) is used to kick the operating 
frequency 100 kHz off the desired 
frequency during receiving periods. This 
prevents interference from the VFO 
while in the receive mode, and enables 
the VFO to remain operational at all 
times, thereby ensuring nearly drift-free 
VFO operation. 

The matching networks and tuned 
circuits of the overall transmitter are 
sufficiently broad in response to permit 
the full 250-kHz operating speed with- 
out need to retune the stages. Rll 
across L4 helps to provide fiat response 
from the VFO chain. 

Circuit-board templates and a parts 
layout are available from the ARRL for 
$1.25 and a large s.a.s.e. 



Fig. 38 - Exterior view of the transmitter 


More Transmitter Topics 51 




Chapter 4 


Power Amplifiers and Matching 
Networks 


P 

Iractical amplifiers and some “cook- 
book" equations will be presented in 
this chapter for those who wish to 
design their own impedance-matching 
networks. Concerning the latter, only 
simple math is needed to solve for the 
various impedance combinations ger- 
mane to solid-state amplifier circuits. It 
is recommended that the builder/ 
designer obtain one of the low-cost 
engineering-function electronic calcula- 
tors for the work treated in this book. 
The resolution is far superior to that 
which can be realized with a slide rule, 
and answers to problems can be ob- 
tained more rapidly with a calculator. 

Despite the large variety of networks 
available for impedance-matching in 
transmitters, all of these designs have 
some common characteristics. First, 
most of the networks used by the 
amateur are essentially low-pass types. 
That is, at frequencies well above the 
design center the networks offer sig- 
nificant attenuation. As a general rule of 



Fig. 1 - Transposition ol a pi network to 
illustrate effect of resistive termination. 


thumb, one can assume that the ul- 
timate attenuation will be 6 dB per 
octave per reactive element in the net- 
work. For example, a common network 
found in the amateur solid-state trans- 
mitter is the double-pi network (low Q ), 
containing two inductors and three 
capacitors. If such a design were “cut” 
for 7 MHz, the attenuation at 14 MHz 
would be around 30 dB. It could be 
higher than this if the network had a 
high- loaded Q. Another characteristic 
of the common impedance-matching 
networks is that they are “singly 
loaded." This fact requires some elab- 
oration: Assume that a low-power trans- 
mitter was being designed for an output 
of 1 watt with a 12-volt dc supply. 
Hence, the required load resistance 
which must be presented to the col- 
lector is V cc 2 -5- 2 P 0 = 72 ohms. A 
suitable network would be a pi type, 
designed to transform a 50-ohm antenna 
termination to the needed 72 ohms. 
What this means is that if one end of 
the network is terminated in a 50-ohms 
resistor, a resistance of 72 ohms is 
“seen" looking into the other end. The 
amplifier behaves as if a 72-ohms re- 
sistor were coupled capacitively to the 
collector. However, the network is not 
being driven from a 72-ohm source. 
Typically, the output impedance of the 
amplifier will be much higher than this, 
perhaps several hundred ohms. 

Networks which are used for im- 
pedance matching are called “singly 
loaded," since it is necessary that only 
one end of the network be properly 
terminated in order to realize the re- 
quired impedance transformation and 
filtering characteristics. Not all LC net- 
works are singly loaded, however. The 
classic double-tuned circuits which one 
mi gilt find in the front end of a receiver 


are doubly loaded designs. That is, both 
the input and output of these networks 
must be terminated properly in order to 
achieve the filtering desired. 

A characteristic of the filters in this 
section is their reciprocal nature. That 
is, even though the networks are singly 
terminated designs, it does not matter 
which end of the network is terminated 
resistively. For example, the pi network 
just mentioned was designed such that a 
50-ohm resistor appears as a 72-ohm 
resistance at the other end. However, 
with the same network a 72-ohm resis- 
tor at the high-Z end would appear as a 
50-ohm impedance at the low-Z end. 
with no difference in filtering prop- 
erties. This is illustrated in Fig. 1. where 
the constants are for 7 MHz and the 
design Q is 3. 

Once the desired resistances for each 
end of a network are determined, the 
network is then “designed.” Inductors 
and/or capacitors are placed either in 
series between the two ends of the 
network, or are connected as shunt 
elements to ground. In the strictest 
sense only two reactive components are 



Fig. 2 — The L network and equations for 

using it. 


52 Chapter 4 






Choose Q (must be grealer Ilian Q 
shown in Pig. 2). 

Then: X,. = QR\ 

*ci =x L -v/rTrT’- r7* 


v _ R1R2 

X C2 ~ y y — 

a l •'Cl 


Fig. 3 - Example of a controlled-O L network 
with equations. 


required to perform any arbitrary im- 
pedance transformation. Such a design 
is realized most directly through the use 
of a Smith chart. 

This simplified approach is some- 
times dangerous, for it leaves the de- 
signer with no control over the Q of the 
network. It a three-element network is 
used, the designer has control over the 
impedance transformation, frequency 
and network Q. Occasionally, one will 
find networks with many additional 
components. The advantage of such 
designs is improved harmonic attenu- 
ation and greater bandwidth. In all of 


r 



Choose Q and Rl greater than R2. 
u _ R1 

Ar i sr- 


\Xc7= R2 


R1/R2 

Q 7 + I RI/R2 


x _ C?RI + Rl R2 /A’c2 
U Q 2 + I 


Fig. 4 - Pi-nelwork configuration with design 
equations. 


the three-element networks described 
next, it is necessary for the designer to 
specify Q at the beginning of the calcu- 
lations. 

The L Network 

This network is a classic for antenna 
matching, but also finds application for 
base and collector matching in solid- 
state transmitters with powers up to a 
few watts. It is not recommended for 
high-power amplifiers. The network is 
shown in Fig. 2 with the design equa- 
tions. Note that R2 must be greater 
than Rl . The Q of the network is given, 
although the designer has no control 
over this parameter. Q is an increasing 
function of the impedance-transforma- 
tion ratio. This accounts for the unde- 
sirability of the network for high-power 
designs. 

The Controlled-0 L Network 

Some of the problems encountered 
with the standard L network can be 
minimized by adding a capacitor in 
series with the existing inductor. A Q is 
first chosen. Then, the equations shown 
in Fig. 3 are applied. 

The Pi Network 

A very familiar circuit is the pi 
network. It has served in the output 
tank of nearly every tube type of 
transmitter built in the last 20 years. A 
wide range of terminations can be ac- 
commodated, including those with sub- 
stantial reactance, and the low-pass 
nature of the network provides excel- 
lent harmonic attenuation. The design 
equations are presented in Fig. 4. 
Manipulation of the equations will show 
that the impedance-matching range of 
the pi network is not unlimited. It may 
be shown that Q 2 + I must be greater 
than Rl 4- R2. For example, a 10-to-l 
transformation is not possible in a net- 
work with a Q of only 2, 

Although useful in some transistor 
circuits, the pi network is not as popular 
as it was in tube-circuit days. The 
primary problem is that the component 
values dictated by the equations are 
sometimes less than practical. For ex- 
ample. it’s not unusual when designing 
an 80-meter transistor transmitter to 
require inductors of 0.5 pH and capac- 
itors of .01 pF. Networks other than the 
pi will lead to more practical com- 
ponent values for die same Q and 
impedance transformation. To general- 
ize, the pi might be best for impedances 
of 50 ohms and higher on both ports. 

The L-C-C Type T Network 

One of the most practical networks 
lor the low impedances common to 
transistors is a T network. It uses a pair 
of capacitors and a single inductor. 
Generally, die component values are F 
practical ii large-value mica-compressior r 



X C2 


OB X, 



Select a Q, 

R2 is greater than Rl . 

Let /I = AR1 (Q’ + I) T 
V R2 ~ 1 

B = R I (Q 2 + I) y C|= . B 
Then X^ = QR\ Q~ A 

X C ? = AR2 


Fig. 5 — The L-C-C matching network with 
related equations. 


trimmers are used. This network is 
limited to the case of R2 being greater 
diap Rl. The equations defining this 
network are given in Fig. 5. The flex- 
ibility of this network is why it is often 
seen in manufacturers' data sheets for rf 
power transistors. 

The L-C-L Type T Network 

If two L networks are combined 
back-to-back, one obtains either a pi 
network or the T network shown in Fig. 6. 
This network has the advantage that 
the component values are often 
practical for solid-state citcuits. How- 
ever, the difficulty in obtaining variable 
inductors with a wide tuning range 
makes the previous L-C-C T network 
more popular. The two-inductor T 
network, nonetheless, offers the ad- 


r 


Choose Q. 

Let A = Rl (C? 2 + 1) 

Then x Ll =RIQ 
*L2 = R2 B 

X c = — A — 
c Q + B 

Fig. 6 - Circuit and equations of the L-C-L T 
network. 


Power Amplifiers and Matching Networks 53 







Fig. 7 - Half-wave filter network circuit. 


vantage of excellent harmonic attenu- 
ation. 

Additional Harmonic Attenuation 

The primary purpose of the net- 
works just presented is the trans- 
formation of impedances. If some of the 
circuits offer superior suppression of 
frequencies above their design center, 
that is certainly a point in their favor. 
However, it should not be a criterion for 
choosing one network over another, for 
harmonic attenuation is easily achieved 
after a transistor has been matched to 
50 ohms. 

A popular method for realizing ad- 
ditional harmonic rejection is adding a 
pi network in the 50-ohms line to the 
load. A convenient network is the 
symmetrical pi (50 ohms, in and out) 
with a Q of 1 . In Fig. 4 we saw that this 
simplified pi section also has easy design 
equations. In this special case, we have 
Xq i = Xq 2 1 Xi = R, where R is the 
termination, usually 50 ohms. If two of 
these filters are cascaded, we have a 
network called the half-wave filler, 
shown in Fig. 7. This name results from 
the properties the network shares with a 
half wavelength of transmission line. 
That is, the phase shift through the 
network is 180 degrees and, more sig- 
nificantly, whatever impedance is used 
to terminate one end of the network is 
the impedance “seen” at the other end. 

Presented in Table 1 are values for 
the components needed to build low- 
power half-wave filters for the amateur 
bands from 1 .8 to 50 MHz. 


Networks like the half-wave filter are 
modified easily to provide infinite at- 
tenuation at specific frequencies higher 
than the design center of the filter. This 
is realized by considering only half of 
the filter of Fig. 7. This symmetrical pi 
network with a Q of unity has the 
design parameters of X c , = X C2 = X^ - 
R. At the design center frequency we 
can modify the filter by replacing any 
of the elements with more complicated 
LC combinations which have the same 
reactance. For example, the inductor 
which has a typical reactance of +50 
ohms at the design frequency could be 
replaced with a trap consisting of a 
parallel LC combination. The behavior 
at the design frequency would be the 
same if the reactance of the series 
element were still +50 ohms. However, 
by properly choosing the components in 
the trap the filter will show virtually 
infinte attenuation at the frequency f„, 
where the trap is self-resonant. The 
design equations for this case are shown 
in Fig. 8. 

Broadband Matching Transformers 

In the preceding section several 
impedance-matching networks were 
presented. One thing a careful observer 
might have noted was that the networks 
would be cumbersome to band switch. 
This difficulty can be avoided through 
the use of broadband matching trans- 
formers. Although these devices have 
appeared frequently in amateur lit- 
erature in connection with solid-state 
linear amplifiers, they may be used 
equally well with Class C amplifiers at 
low or high power levels. Like the 
narrowband networks of the previous 
section, broadband transformers may be 
considered as singly terminated re- 
ciprocal networks. 

Of the broadband rf transformers 
there are basically two types. One is 
essentially a conventional transformer 
which has been adapted for the low 
impedances common to high-power 
amplifiers (more on these transformers 



Fig. 8 — Modification of the half-wave filter to 
provide added harmonic attenuation. 



Fig. 9 — Principles of an ideal transformer, 
with waveforms. 



Fig. 10 — Illustration of current flow in a 
bifilar-wound transformer. 


54 Chapter 4 







i 


•LuuJ 

«L 

l «pnnr| f 



"SORTABALUN" 



Fig. 1 1 — Circuit for an isolation transformer. 


later in this chapter). The other con- 
figuration is the broadband trans- 
mission-line transformer. These trans- 
formers act as conventional trans- 
formers at their lower operating fre- 
quency, but act as transmission lines 
near their upper frequency limit. To 
attempt a complete explanation would 
be beyond the scope of this presen- 
tation. Hence, we will provide an over- 
view and a few rules of thumb for the 
construction of the transmission-line 
transformers. 

It is well known that a quarter 
wavelength of transmission line exhibits 
impedance-transformation properties. If 
a X/4 length of line with a characteristic 
impedance Z a is terminated with a 
resistance Rl, a resistance R2 is seen at 
the other end of the line, where Z„ 2 = 
R1R2. For example, if a 35-ohm resis- 
tive load, such as the base of a ground- 
plane antenna, is placed at one end of a 
X/4 length of 52-ohm coax cable, a 
resistance of 77 ohms is presented at the 
other end, offering a good match for 
RG-1 1 cable. The same principles apply 
for other kinds of lines (in this case, 
twisted pairs of insulated wire). Al- 
though the pitch of the twist can have 
some effect on characteristic impedance 
of the line, as does the wire diameter 
and insulation thickness, we will ignore 
these effects for the most part. One can 
assume generally that a twisted pair of 
plastic-covered hookup wire will have a 
Z a of about 100 ohms. Similarly, a 
twisted pair of No. 24 enameled wires, 
twisted to about five turns per centi- 
meter, will end up near 50 ohms. 
Twisted pairs are formed easily by 
clamping one end of the pair in a vise. 
The other end is hooked through a 



Fig. 12 - Circuit of a 4:1 step-up transformer. 


“fishhook" formed from large-diameter 
wire which is inserted in the chuck of a 
hand drill. With the wire held taut, the 
drill is operated until the proper pitch is 
obtained. 

Twisted pairs could be used directly 
for transmission-line transformers ex- 
cept for a couple of problems. First, a 
quarter wavelength of line at, say, 80 
meters is less than practical. This is 
where a toroid core comes in. The 
second problem is that the impedances 
usually needed for solid-state power 
amplifiers often dictate the use of low- 
impedance transmission lines, with Z a 
well below 50 ohms. For example, an 
amplifier designed for an output of 6 
watts from a 12.5-volt dc supply would 
require a load resistance of 12.5 ohms. 
A 50-ohm output termination could be 
transformed to 12.5 ohms by a line of 
Z 0 = 25 ohms. This 25-ohm line is 
realized easily by paralleling two 50- 
ohm lines. Often, for the really low 
impedances needed for base matching, 
the required low-impedance lines are 
formed by paralleling as many as four or 
five line pairs. 

We will now depart momentarily 
from our consideration of transmission 
lines and review the behavior of an ideal 
transformer. Consider first the relatively 
simple case of a single inductor, for 
example, a winding on a ferrite toroid. 
Recall that an inductor is a component 
in which the current flow cannot change 
instantaneously. If our hypothetical 
inductor is connected directly to a 
battery, the voltage across the battery 
immediately appears across the in- 
ductor. However, the current flowing in 
the inductor is initially zero: After all, 
the current was zero prior to application 
of the battery. The waveforms are 
shown in Fig. 9 along with the circuit. 
The fact that the current builds up 
slowly is a result of the changing mag- 
netic field in the core. This changing 
field induces a voltage across the coil 
which impedes the flow of a net cur- 
rent. The current in the coil will, how- 
ever, grow in time, leveling off at the 
level dictated by the internal resistance 
of the coil and of the battery. If we had 
ideal components with no internal resis- 
tance, the current would grow linearly 
forever. 

Consider now the bifilar-wound 
transformer shown in Fig. 10. Again, we 
connect a battery to the primary of this 
transformer. In this case, however, cur- 
rent can flow instantaneously. As soon 
as the smallest current begins to flow in 
the primary, AA , the resulting magnetic 
field causes a voltage to appear across 
the secondary, BB'. This voltage causes 
a current to flow through the resistor 
loading the secondary. This secondary 
current, in turn, establishes a magnetic 
field which opposes the field caused by 
the current in the primary. Hence, with 



| A A* 

21 

4R ► 


* >» 


^ * rrrr b- 7 

I 


4 1 STEP DOWN 



Fig, 1 3 - A 4 : 1 transformer which has fre- 
quent use in collector matching. 


a net magnetic field of zero, there is no 
inductive voltage to oppose current flow 
in the primary. The current flow is 
exactly the same as if the resistor were 
connected directly to the battery. 

Since transformers work only on 
changing magnetic fields, the trans- 
former will eventually cease to work 
when the core saturates. However, with 
ac signals, such as the rf of our present 
concern, the fields are always changing 
at rf rates. 

It is important to note the direction 
of current flow and the dots in the 
figure which indicate voltage polarities. 
That is, a positive-going voltage applied 
at one dot will lead to a positive-going 
voltage at the other dot. The directions 
indicated for instantaneous current flow 
are those required for transformer ac- 
tion. 

It is instructive to consider some of 
the transformer configurations which 
are of practical utility in rf design. Only 
some of the more straightforward types 
will be presented. Shown in Fig. 11 is an 
isolation transformer. This configura- 
tion is often called a balun, although it 
does not really deserve this name, for 
the transformer does not force the 



Power Amplifiers and Matching Networks 55 






Fig. 15 - Illustration of a 9:1 unbalanced 
transformer. 


voltage applied across the resistor, R, to 
be balanced with respect to ground. 
Indeed, if the end of the resistor con- 
nected to the primary were grounded, 
the input voltage would appear across R 
except that there would be a phase 
reversal. On the other hand, if R con- 
sisted of a pair of resistors in series, with 
their junction grounded, the input volt- 
age would appear as a balanced, equal 
voltage across the balanced load. Be- 
cause of the similarity to a balun trans- 
former, WA6RDZ has suggested that 
this configuration be called a “sorta- 
balun.” Note in Fig. 1 1 that the current 
in the transformer is in the proper 
direction to preserve transformer action. 
However, any voltage common to both 
leads at one end of the sortabalun or the 
other will see a very high inductance, 
with minimal resulting current flow. 
Hence, the excellent isolation prop- 
erties. 

Presented in Fig. 12 is a 4:1 step-up 
balun (for real) transformer. The trans- 
former is drawn in two different ways 
to emphasize the variety of approaches 
one can use in the analysis of such 
components. The sketch at A shows 
that the drive voltage is applied across 
one winding of a center-tapped coil, 
with the termination across both parts 
of the coil. The diagram at B emphasizes 
the direction of current flow which 
must exist for proper transformer 
action. Clearly, there is twice as much 
current flowing from the source as that 
flowing in the resistive load, implying a 
4:1 impedance transformation. 

Several of the other transformer 
configurations are presented in Figs. 13 



Fig. 16 — A 4:1 balanced-to-balanced 
transformer. 


through 16. The 4:1 type in Fig. 13 is 
commonly used for collector matching 
in medium-power amplifiers. Two trans- 
formers of this kind may be cascaded 
for a 16:1 transformation for matching 
50 ohms to the base of a high-power 
stage. The 1:1 balun transformer (Fig. 
14) is often used with balanced antenna 
systems. Note that this is a, real balun 
rather than a “sortabalun.” 

The last two figures show trans- 
formers which use two toroid cores. The 
9:1 single-ended configuration is useful 
for base matching in medium-power 
amplifiers. The 4:1 balanced-to- 
balanced configuration is sometimes 
used with push-pull high-power am- 
plifiers. Typically, this 4:1 transformer 
is combined with an isolating sortabalun 
at the end, which must ultimately be 
terminated. Point “X” may be grounded 
if it is necessary to force balance. With 
care, either the 9:1 or 4:1 transformers 


can be wound on single cores. The 
reader is referred to Motorola Applica- 
tions Note AN 593 for this subject. 

Little has been said about the con- 
struction of practical versions of the 
transformers we have discussed. 
Fortunately, building them is straight- 
forward. The first step is to obtain 
suitable toroids. Most of the toroidal 
cores used in amateur radio are of 
powdered iron and are used in tuned- 
circuit applications. However, for 
broad-band transformers, ferrite cores 
are preferred (p of 125 to as great as 
950). The main reason for this is that 
ferrite exhibits a much higher permea- 
bility than most of the powdered-iron 
cores used in the hf region. Because of 
the high initial permeability, the in- 
ductances required for good transformer 
action are realized with a minimum 
number of turns. This minimizes prob- 
lems with self-resonances in the cores. 
Both ferrite and powdered-iron cores 
are available from Amidon Associates 
(see QST ads). 

The next step is to consider the 
transmission-line requirements of the 


transformer. As mentioned earlier, at 
the higher frequency end of the op- 
erating range of most of these trans- 
formers the core has minimal effect. It 
is the transmission line which performs 
the desired transformation. The core is 
of significance only at the lower fre- 
quencies. The required end impedances 
of the transformer are first determined. 
For example, a 4:1 transformer for 
collector matching in the 6-watl am- 
plifier mentioned earlier must match 
between 50 and 12.5 ohms. The re-, 
quired 7, a is given by VRI R2, or in this 
case, 25 ohms. This is obtained best 
with two paralleled 50-ohm lines. 

It is helpful to make each twisted 
pair from enameled wire of two dif- 
ferent colors. If this is not possible, it 
might be worthwhile to paint one of the 
wires with a suitable coloring agent, or 
tie a knot at each end of one wire. The 
twisted pair is made with a hand drill as 


outlined earlier. Then, two of these 
pairs are paralleled and twisted loosely 
with the drill. In this case, a couple of 
twists per centimeter is probably more 
than sufficient. This bundle of four 
wires is then wound through the core 
several times. The accepted rule of 
thumb is that the length of the winding 
should be 1/8 wavelength at the highest 
operating frequency, although much less 
wire will often work satisfactorily. 
Then, after winding, the ends of the 
wires are stripped of insulation. As- 
suming the two colors are red and green, 
the beginnings of the two red wires are 
twisted together as are the beginnings of 
the two green wires. The ends of the red 
and green wires are treated in a similar 
fashion. Having four wires now. we can 
assign the green wire as "A” and the red 
wire as “B" and wire the transformer as 
shown in Fig. 13. For transformers with 
lower characteristic impedances, similar 
procedures are followed with, of course, 
more than two paralleled twisted pairs. 

Several transformers have been built 
and studied with a network analyzer. In 
both cases to be described, the toroids 



56 Chapter 4 





,/ 2 -r m 



0 .2 4 6 e 1.0 


0CS-THERMAL RESISTANCE -CASE TO HEAT SINK 


Fig. 18 - Representation of the thermal 
resistance of a transistor case to the heat sink 
(see text). 


had an initial permeability of 125, and 
had an OD of 0.375 inch (Amidon 
Associates FT-37-61). The first case 
studied was a 4:1 transformer suitable 
for the output of a 25-watt amplifier 
with a 24-volt supply. Three turns of 
two bifilar pairs of No. 24 enamel wire 
were wound on a stack of four of the 
toroids. The high -impedance end of the 
transformer was terminated in 50 ohms, 
and the input impedance of the low- 
impedance port was measured. In 
scanning the range from 3.5 to 21 MHz, 
the measured impedance varied from 
12.5 + /3 .6 to 13.3 + /4.3. The slightly 
inductive impedance seen should 
present no problem in an amplifier, for 
the transistor is slightly capacitive. 

The second case studied was a com- 
posite 16:1 transformer formed from 
two 4:1 transformers. The first trans- 
former (50 ohms to 12.5 ohms) used 
one core wound with six bifilar turns of 
one twisted pair of No. 26 wire. The 
second used two twisted pairs on a 
single core, again only six turns. By the 
rules outlined above, the first core 
should have used two twisted pairs, and 
the second should have had eight! The 
cores, however, were too small to accept 
this much wire. In spite of the departure 
from the design ideals, the 16:1 trans- 
former looked reasonable, although still 
inductive. With the high-impedance end 
of the composite transformer ter- 
minated in 50 ohms, the impedance 
seen at the other port ranged from 3.4 + 
/ 1.4 at 3.5 MHz to 3.2 + /4.5 at 21 MHz. 
The relatively high reactance would 
probably require some capacitive com- 
pensation at the higher frequencies. 

A medium-power cw amplifier was 
breadboarded using the two trans- 
formers described above, and is shown 
in Fig. 17. The transistor used was a 
Motorola 2N5942. This device is spec- 
ified for 80-watts PEP linear output, so 
it was loafing in the 25-watt test circuit. 
Nonetheless, the performance was just 
about that expected when tested at 7 


and 14 MHz. An output of 25 W was 
obtained easily on both bands with a 
24-volt power supply. The drive re- 
quired on 20 meters was about 0.5 watt, 
while 250 mW were sufficient on 40 
meters. No instability problems were 
noted. 

High-Power Solid-State Amplifiers 

There was a time when transistor 
transmitters were for low -power enthu- 
siasts. It was not a matter of choice — 
the only transistors available were low- 
power devices. Today final stages with 
an output of 100 watts or more are 
practical and economical. In a few years 
the amateur may no longer be able to 
purchase a transceiver in this power 
class with even a single tube in the 
circuit. Through the use of hybrid- 
power splitters and combiners, a 
number of amplifiers in the 100- to 
300-watt output class have been com- 
bined to yield over I kW of output. 
Most of the problems encountered in 
building a high-power amplifier are sim- 
ilar to those outlined earlier for low- 
power stages. 

Almost all modern rf power devices 
are specified for operation in the fre- 
quency range for which they were de- 
signed. Most manufacturers’ data sheets 
include curves of input resistance, input 
reactance and output capacitance as a 
function of frequency. Output load 
resistance is not often specified, since 
the equation (R L = V cc 2 -f 2 P p ) is 
sufficiently accurate. With transistors 
specified for the hf region, most of the 
data are for linear operation. However, 
the information is close enough for use 
in designing Class C stages for cw and 
fm. 

Heat Sinking and Mounting 

The main difference between a high- 
power amplifier and one for QRP work 
is the level of heat sinking required. The 
efficiencies quoted by manufacturers 
vary, but a ball-park number might be 
65 percent for Class C service, and 30 to 
50 percent for Class AB or B linear 
amplification. The builder should 
expect that as much power will be 
dissipated in heat as will be obtained in 
rf power output. Certain prescribed 
methods should be followed to ensure 
long transistor life, as heat in excessive 
amounts (junction temperature) is one 
of the major enemies of power tran- 
sistors. 

The thermal resistance (resistance of 
a material to heat transfer) from a 
transistor case to the heat sink is any- 
thing but incidental. Fig. 18 shows 
typical values of thermal resistance for 
different package types when the de- 
vices are bolted to their heat sinks in 
accordance with the manufacturers' 
specified torque. The latter is usually 6 
±l-inch pounds for 3/8-inch studs, 5 



Fig. 19 — Correct and incorrect mounting 
methods for stud transistors with strip-line 
connector leads. 


TRANSISTOR 

JUNCTION 


TRANSISTOR JUNCTION TO CASE 
’ R1(0|c) THERMAL RESISTANCE 

SPECIFIED BY TRANSISTOR 
MANUFACTURER 


» CASE TO HEAT SINK 

•R2<0cj> thermal resistance 


► LATERAL-HEAT- TRANSFER-TO- 
< R . FINNED AREA-OF-HEAT SINK 

> THERMAL RESISTANCE. USUALLY 
SPECIFIED AS ONE TERM BY HEAT 
SINK MANUFACTURER 


HEAT SINK-FINS-TO-AIR 
THERMAL RESISTANCE 


Fig. 20 — Resistances to heat flow when a 
transistor is joined to a heat sink. 


Power Amplifiers and Matching Networks 57 






± 1 -inch pounds for 1/4-inch studs, and 
8 ± 1 -inch pounds for 1/2 -inch studs, 

Thermally induced mechanical stress 
should not appear anywhere in the 
transistor. It is for this reason that 
correct torque is important. Further- 
more, the surface of the heat sink to 
which the transistor case mates must be 
as smooth and flat as practicable. A thin 
layer of heat-transfer silicone grease 
should be coated on the stud and 
interface portions of the transistor and 
heat sink prior to mounting. 

Strip-line types of transistors (wide, 
flat emitter, base and collector external 
leads) should be mounted so that the 
leads are not stressed. Furthermore, the 
circuit-board foils to which they con- 
nect should be brought as close to the 
transistor body as possible to prevent 
unwanted inductances from being 
formed by the strips (Fig. 19A). When 
the leads are bent as shown in Fig. 19B 
and C, stress exists, and may increase 
when heating occurs. A bad effect from 
bent emitter strips is that of degenera- 
tion caused by the excessive inductance 
which results. This will lower stage gain, 
and is a particularly significant matter as 
the operating frequency is increased to 
the upper hf region, and at vhf and uhf. 
Fig. 20 shows the resistances to heat 
flow which occur when a transistor is 
joined to a heat sink. 

If the foregoing ideal guidelines can't 
be followed, the amateur can use the 
following procedure to assure safe oper- 
ation. Start by bringing the power sup- 
ply voltage up slowly, and monitor the 
collector current continuously. Make 



Fig. 22 - Example of a recommended pc- 
Ooard foil panern for use with stud-mouni 

strip-line transistors. 


frequent checks of the transistor and 
heat-sink temperatures by touching a 
finger to each element. If the transistor 
body becomes too hot to endure with 
comfort, excessive heat will be present. 
This will indicate that the heat sink is 
not of adequate area, that thermal 
bonding is improper, or that excessive 
collector current is flowing. If a torque 
wrench is not available, tighten the stud 
nut just beyond the point where it is 
finger tight. Transistor mounting and 
heat considerations are treated in 
Motorola Application Note AN-555, 
and in Solid Circuits by Communi- 
cations Transistor Corp. of San Carlos, 
CA. 

It is not necessary to purchase heat 
sinks if aluminum sheeting is available. 
Large heat sinks can be fashioned from 
U-shaped pieces of heavy-gauge alum- 
inum. as shown in Fig. 21. Homemade 
sinks are inexpensive and can be put to 
use quickly. 

The use of wide pc-board foils is 
recommended in rf portions of the 
circuit. Wide foils will lessen the un- 
wanted inductance effects, and will 
make soldering of the transistor strip 
lines easier. An illustration of the prin- 
ciple is given in Fig. 22. Double-sided 
pc-board material (copper on both 
sides) is almost mandatory in the in- 
terest of electrical stability. The side 
opposite the foils and transistor body 
serves as a ground-plane surface to 
discourage current loops which can 
cause feedback. Additionally, the 
ground plane acts as one plate ot a 
capacitor for each of the etched foils, 
affording vhf and uhf bypassing 
throughout the board. This also helps 
prevent unstable operation. The 
ground-plane side of the board should 
be made electrically common to the 
ground foils on the etched side of the 
board. 

Some Electrical Considerations 

It is practically impossible to lay 
down a definite rule for selecting a 
power transistor which must deliver a 
specific output power. Conuncrcial de- 
signers have, on occasion, pushed power 
transistors quite hard - extracting 
power amounts which were as great as 


3/4 P D maximum. That is, a transistor 
with a miximum safe power-dissipation 
of I0W at 25°C might be called upon 
to deliver 7 watts of rf output when 
installed on an adequate heat sink with 
correct mounting techniques. In ama- 
teur work that kind of courage is not 
recommended. A transistor operated 
within sensible ratings should last for 
100,000 hours of "on" time, at the 
least. That kind of longevity would not 
be typical of an amateur amplifier if it 
were "milked" for all it was worth. A 
good rule of thumb is to select a 
transistor which has a Pp(,„ ax i of 
roughly twice the power it will deliver. 

It is not especially wasteful of money 
and device capability to make the safety 
margin even greater. When more power 
output is needed than the Pp rule of 
thumb can assure, use a larger single 
transistor, or two in push-pull, instead 
of paralleling two smaller ones. This will 
reduce cost somewhat, and will make 
the circuit less difficult to optimize. 
When two or more devices are used in 
parallel, layout and load-sharing prob- 
lems become difficult to predict and 
control. 

It is not recommended that vhf or 
uhf transistors be used in mf and low-ht 
band power amplifiers. The gain (Fig. 
23) increases markedly as the operating 
frequency is lowered (6 dB per octave), 
and this can make stabilization ex- 
tremely difficult. It is best to utilize 
transistors which were designed for the 
frequency range of interest. Further- 
more, a power transistor should be 
operated at a power-output figure which 
is 75 to 80 percent of the saturated 
power output. That approach will assure 
best efficiency and will reduce power 
drop-off with heating. (Saturated power 
output is that point where further out- 
put can’t be obtained with increased 
drive.) 

Gain Compensation 

Broadband amplifiers require gain 
equalization if a wide range of fre- 
quencies must be accommodated, say, 
1.8 to 30 MHz. It was said earlier that 
transistors have increasing gain at ap- 





52 

46 

40 

34 

28 

o 

22 

46 

40 

4 

















































e 5 

.5 7 14 21 28 56 

MHi 


Fig. 23 - Curve showing the 6-dB-per-octave 
gain characteristic of a transistor. 


58 Chapter 4 











Fig. 24 - Gain-compensating network, 
labeled L and R. 


proximately 6 dB per octave lower, 
which means that very high gain is 
probable at the low end of the amplifier 
range. It is desirable to equalize am- 
plifier gain as much as possible to 
prevent the necessity for a variable 
drive-power exciter, and to prevent 
damage to the transistors from parasitic 
oscillation or excessive collector current 
at the low end of the operating range. 
Two forms of compensation are 
popular, and each requires some em- 
pirical adjustment. 

One technique is to add a "losser” 
network at the input to the amplifier 
(Fig. 24). Inductance I. is selected to 
have low reactance in the range where 
the gain increase is significant, and as 
the operating frequency is made lower, 
the loss through the compensating net- 
work increases. Addition of resistance R 
serves a twofold purpose — it lowers the 
network Q and provides a load for the 
driving power that must be dissipated 
external to the transistor. It is some- 
times necessary to add component C to 
correct for a mismatch caused by the 
compensating network. Depending on 
the capability of the exciter with re- 
spect to SWR, a moderate amount of 
mismatch may be tolerable at the low 
end of the amplifier frequency range. 
Generally, input SWR should be made 
lowest at the high end of the amplifier 
operating range. 

Another technique used by some 
designers to equalize amplifier response 
is to employ negative feedback (col- 
lector to base). The method is il- 
lustrated in Fig. 25 for a single-ended 
amplifier, and in Fig. 26 for a push-pull 
module. The principle is one of adding 
an R-C network which has the property 
of increasing the negative feedback as 
the operating frequency is lowered. The 
component values depend on the device 
characteristics, power levels and im- 
pedance characteristics of the amplifier. 
Therefore, no set rules for component 
values are offered here. (See chapter 8 
for details.) Typical values for a 50-W 
amplifier might be 10 ohms and 100 pF 
for the circuits of Figs. 25 and 26, 
assuming an amplifier bandwidth of 1 .8 
to 30 MHz. It should be said that any 


compensating network a builder may 
add to an amplifier will have some 
effect on the circuit, and caution should 
be used when such L-C-R sections are 
included in a design. 

Ballasted Transistors 

Modern power transistors for linear- 
amplifier service are emitter ballasted. 
That is, each emitter or group of emit- 
ters in a device (several bipolar tran- 
sistors are used in parallel on a single 
substrate) contains a separate series re- 
sistance. This feature helps prevent hot- 
spotting on tire chip (second break- 
down) which can occur anywhere on 
the complex internal surface. Hot 
spotting takes place when one or more 
of the individual transistors on the 
substrate ‘ hog" power. The result is 
failure of the composite transistor. The 
series resistances tend to equalize the 
current sharing as changes occur ex- 
ternally, thereby protecting the tran- 
sistor from damage. The possibility of 
second breakdown is related mainly to 
linear transistors (but also affects Class 
C amplifiers) because forward bias is 
applied. Therefore, when SWR is high, 
or when strong self-oscillation takes 
place, hot-spotting is likely to become 
manifest. Ballasted transistors are ex- 
cellent for all classes of operation - A, 
AB, B and C. 

A protective measure for unballasted 
transistors is seen in Fig. 27. A Zener 
diode is connected as a peak-voltage 
clamp from collector to ground. As- 
suming the maximum collector voltage 
swing will be twice the supply amount 
(24 V), VR1 is not pait of the collector 
circuit. However, should a load mis- 
match occur, or the stage break into 
self oscillation, the collector rf voltage 
will soar to high value. At that point 
VR1 will conduct at 36 V and clamp 
the voltage above that value, thereby 
protecting the transistor. Furthermore, 



Fig. 25 - Negative-feedback gain compensa- 
tion using C and R components. 


should voltage spikes occur on the 
supply line the Zener diode will clamp 
at 36 V or higher again protecting the 
transistor. If protection against ex- 
cessive positive and negative voltage 
swings is desired, two Zener diodes can 
be bridged from collector to ground, 
back-to-back fashion. ARRL lab tests 
indicate that no degradation in amplifier 
performance results from use of Zener- 
diode clamps at hf and mf, provided the 
diode conduction point is well above 
the normal rf-voltage peak value. No 
evidence has been found that VR1 
enhances the generation of harmonic 
currents while in its “off’ state. 

Protective measures should be as- 
sured for any piece of solid-state equip- 
ment which operates from a dc supply 
that is not treated for transient sup- 
pression. Notably, mobile gear which 
uses the automotive ignition supply for 
operating voltage can be subjected to 
large voltage spikes that can ruin the 
transistors or ICs. A good safety pre- 
caution is to add an 18-V, 10-W Zener 
diode from the 12-volt input line to 
ground. The same principle applies to 
equipment which is powered by ac- 
operated dc supplies that have no spike- 
protector circuits. 

Conventional Broadband Transformers 

Considerable treatment was given 



Power Amplifiers and Matching Networks 59 







Fig. 27 - Example of a Zener -diode protective clamp at the collector of a power amplifier. 


earlier to the design and use of trans- 
mission-line transformers for broadband 
applications. It is worth mentioning that 
conventional broadband transformers 
are also suitable for many amateur 
circuits. A number of commercial man- 
ufacturers are using conventional trans- 
formers in their power blocks, and with 
good results. 

Most broadband rf transformers of 
tire “conventional” type are toroidal 
and use iron or ferrite cores. However, 
ferrite rods can also be used as the core 
material in conventional broadband 
transformers. The self-shielding prop- 
erties of toroid cores are preferable in 
most amateur work, however. 

Fig. 28A shows the electrical rep- 
resentation of a conventional broadband 
transformer. LI is a small-diameter brass 



Fig. 28 - Circuits of a conventional broad- 
Band transformer with sketch of how they 
are constructed (see text). 


or copper tube which is U-shaped, and 
over which several high-/a toroid cores 
have been placed (permeability = 950 in 
most designs for mf and hf)- The ends 
of the tubes are soldered to the pc- 
board plates as shown at B. U-shaped LI 
functions as a 1-turn secondary winding, 
and is hooked to the bases of push-push 
amplifier transistors when T1 is used as 
an input transformer. Alternatively, T1 
can be used as an output transformer, in 
which case, the ends of LI connect to 
tire collectors of the amplifier, or to the 
balanced winding of a collector rf 
choke. LI establishes the turns ratio of 
the transformer by virtue of its being a 
1-turn winding. Insulated hookup wire 
is passed through the tubes of LI and 
serves as the primary winding (L2) of an 
input transformer, or as the secondary 
of an output transformer. The number 
of turns used will depend upon the 
impedance-transformation ratio needed. 

The number and size of the ferrite 
cores used will be related to the power 
level of the amplifier and the desired 
reactance of the windings. A good rule 
of thumb is to make the transformer 
windings exhibit four to five times the 
impedance of the circuit to which the 
transformer is connected. Thus, a 
winding that connects to a 50-ohms 
load should look like, say, 250 ohms at 
tire lowest operating frequency. 

One advantage of the conventional 
transformer of Fig. 28 (and many trans- 
mission-line transformers) is that ex- 
cellent symmetry results from the 
construction style, and symmetry is 
essential when obtaining electrical 
balance in push-pull power amplifiers. 
The pc -board end plates of the trans- 
former can be soldered directly to the 
main pc -board pads to which they re- 
late. A photograph of some con- 
ventional broadband transformers is 
shown in Fig. 29. 

Other Considerations 

Occasionally, the amateur will use 
transistors which have the fr and power 
capabilities for rf-power applications, 
but lack the specifications needed for a 
really complete “paper” design. These 
devices can often be used for amplifiers 
by making reasonable estimates of the 



Fig. 29 - Photograph of some conventional 
and transmission-line transformers. The 
unit with the twisted wires (center) is a 
transmission-line transformer. 


parameters. Some of the guessing pro- 
cedures will be outlined without de- 
tailed justification. First, the f T of the 
device being well above the operating 
frequency (a factor of three, four, or 
more) will ensure that a reasonable 
power gain is available. The Vceo of 
the device should exceed the operating 
voltage, V cc , by a factor of two or more 
in cw and linear applications. Ideally, 
the beta of the transistor should hold up 
well at the desired collector current. If 
these criteria are met, the operating 
parameters are easy to guess. The out- 
put resistance needed is V cc 2 -r 2 P„. 
Usually, the output capacitance ( C a ) 
can be ignored: It can be absorbed in 
the output tuning network of a narrow- 
band design. BroadtV.d designs may 
present more problem^ however. The 
input resistance is related to the current 
grin at the frequency of operation and 
is inversely related to the output power. 
For amplifiers in the 20- to 70-watt 
output region, one can arrive at a 
satisfactory design by assuming an input 
resistance of around 2 ohms. If an L-C-C 
type of T network is used for matching, 
with a design Q of 5, input resistances 
of less than 1 ohm may still be accom- 
modated without excessive network Q 
(see Fig. 5). It is possible to neglect the 
input reactance of the base, allowing the 
reactance to be absorbed in the im- 
pedance-transforming network. As a 



Fig. 30 — Circuit of the modified L-C-C 
network. 


60 Chapter 4 






conservative rule of thumb, one should 
never design for an output power ex- 
ceeding die heat dissipation of the 
transistor being used. Less is a better 
and safer assumption. 

As was outlined earlier, there is a 
wide variety of networks from which to 
choose for impedance matching. How- 
ever. the L-C-C type of T network is an 
excellent first choice for base and col- 
lector ma tching, owing primarily to the 
range of impedances which may be 
accommodated with a given network 
design, and to the practicality of the 
component values. It is worthwhile to 
modify the output network slightly by 
adding some additional capacitance in 
parallel with the collector. A reasonable 
value is a reactance of two or three 
times the output resistance, Fig. 30. 
This added capacitance will have little 
effect at the design frequency, but will 
significantly aid in the suppression of 
vhf parasi tics. This is of major signif- 
icance it a vhf power device is used in 
the hf region. 

One can pretune the networks to the 
design frequency and impedance before 
power and drive are applied to an 
amplifier. This prealignment is done 
easily with a 50-ohm impedance bridge 
and a low-level rf source. (A suitable 
bridge is described in a later chapter.) 

As an example, assume that an amplifier 
will deliver an output of 50 watts with a 
28-volt power supply. The collector 
load resistance will be V cc 2 ±2P 0 =7.8 
ohms. The network is designed and a 
reasonably close-value resistance is con- 



F>9. 32 - Circuit of the input part ot an 
amplifier. 


nected temporarily to the circuit as 
shewn in Fig. 31. In this example, an 
appropriate resistance would be a pair 
of paralleled 15-ohm carbon resistors. 
The network is adjusted for a bridge 
null, indicating that 50 ohms exists at 
the output port. The 7-1/2 ohm resistor 
is then removed from the circuit! 

Shown in Fig. 32 is the input part of 
a power amplifier. The rf choke serves 
as a dc path for the flow of base 
current. Since the input resistance of 
the transistor is very low, the reactance 
of tins choke is not critical and is 
usually four or five times the input 
resistance. However, the Q of this choke 
should be quite low, often less than I. 
This is realized by shunting the choke 
with a low-value resistor, less than the 
reactance of the choke. Even lower 
values (down to an ohm or two), 
comparable to the value of the tran- 
sistor input resistance, will add to the 
stability of the amplifier. If this practice 
is followed, the input network may be 
prealigned with a bridge without substi- 
tution of extra base resistance. 

Once an amplifier is built and pre- 


aligned, the moment of truth comes 
when dc power and rf drive are applied. 
Tlie output is terminated in a 50-ohm 
resistive load with means for measuring 
power output. The light bulb load of 
the tube era has no place in the modern 
amateur lab. and should not be used as 
an rf termination! A current-limited 
power supply should be used. Initially, 
the voltage is reduced to half of the 
normal operating level in die case of 
high-voltage amplifiers (e.g., 28 volts). 
For stages operating from 12 volts it is 
suitable to begin experimentation at 
that level. A low amount of rf drive is 
applied and the output is noted. The 
networks are adjusted for maximum 
output, always keeping an eye toward 
signs of instability. This procedure is 
repeated at increased power-supply volt- 
ages and rf drive levels, keeping the 
networks tuned for maximum power 
output. The collector current should be 
monitored for any tendency toward 
thermal runaway, and the device and 
heat-sink temperature should be mon- 
itored. 

If the amplifier has forward bias, as 
is typical of linear amplifiers, careful 
attention should be devoted to moni- 
toring the current during application of 
rl drive, and afterward. Many amplifiers 
which perform well in ssb service may 
not be capable of withstanding the 
tremendous power dissipation levels in- 
curred during cw testing or two-tone 
evaluation. 

A final problem which can occur 
with high -power amplifiers should be 
mentioned. Often the collector current 
in a high-power amplifier is several 
amperes. With such a high current it can 
be extremely difficult to decouple the 
amplifier Irom the remaining circuit. 
Additional decoupling networks may be 







General-purpose 6-watt rf amplifier which uses a single transistor. The amplifier is seen at the 
bottom of the photograph. This WA7MLH unit contains half-wave output filters for 80 and 
40 meters, plus a small relay which, when actuated, bypasses the amplifier for QRP operation. 


required. In the home station it is 
worthwhile to operate a high-power 
final stage from a power supply separate 
from that used to power the rest of the 
station. 

Broadband Utility Power Amplifiers 

Many QRP transmitters built by the 
experimenter have an output of a watt 
or less. The amplifiers shown in Figs. 33 
and 34 are designed to complement 
such rigs, providing outputs of four to 
six watts, while not presenting a strain 
on the pocketbook. Both designs use 
broadband matching transformers of the 
type outlined in a section earlier. They 
are suitable for the amateur bands up 
throu^i 20 meters. 

The simpler of the pair of amplifiers 
(Fig. 33) has a single-ended design using 
one transistor. All three transformers 
are wound identically. T1 and T2 are 
wired as a composite 9:1 step-down 
transformer such that the base of the 
transistor is driven from a source of 
approximately 6 ohms. The output re- 
sistance of 12 ohms is matched to a 
50-ohm termination with T3, which is 
wired as a 4:1 step-up. 

Several transistors were tried in the 
single-ended configuration. Excellent re- 
sults were obtained with the GE D446C, 
which is available for just over SI. This 
device has an /y of 50 MHz, a 30-watt 
collector dissipation, and a V C eo of 45 
volts making it ideal for rf-power ap- 
plications on the lower bands. With this 
transistor, output powers of 6 watts 
have been obtained on 80 meters, with 


4-1/2 to 5 watts being more typical for 
40 meters. The power gain is roughly 10 
dB on 40 meters. It approaches 16 dB at 
3.5 MHz. Versions of this amplifier have 
been used by West Coast amateurs for 
tire output of QRP transceivers which 
were designed specifically for Field Day 
use. Since the efficiency is about 50 
percent, the amplifiers are ideal for the 
10-watt input limit in the QRP cate- 
gory. The RCA 2N5321 is worth invest- 
igating as a substitute in this circuit. 


Shown in Fig. 34 and in the photo- 
graph is a push-pull version. Although 
slightly more complicated than the 
single-ended amplifier, this scheme is 
worth consideration. First, the push-pull 
version has the advantage of twice as 
much power dissipation. Furthermore, 
even-order harmonics are suppressed by 
the balanced circuit. Finally, and this is 
of significance when using inexpensive 
transistors not intended for rf power 
application, a higher output-load resis- 
tance may be used. This allows reason- 
able efficiency to be maintained without 
requiring that the transistors have good 
saturation specifications. 

In the push-pull amplifier Tl steps 
the input 50-ohm drive down to 12.5 
ohms in a single-ended manner. T2 then 
provides drive to the balanced bases. 
The third core in the input section of 
die amplifier ensures that the load 
presented to T2 is balanced. Each tran- 
sistor sees a driving impedance of 6-1/4 
ohms. In chapter 2 it was noted that at 
low frequencies a problem sometimes 
encountered with power stages is break- 
down of die emitter-base diode of the 
transistor. The use of push-pull circuitry 
prevents diis from happening, for each 
transistor acts like a negative-clamping 
diode for the other. 

In die output T4 plays two roles. 
First, it provides a path for the dc bias 
to reach the collectors. Saturation of 
the toroid is no problem at high cur- 
rents, since the current dirough the 
opposing windings sets up opposing 
fields. The second purpose of T4 is to 
ensure dial the collectors are balanced. 
T5 transforms die balanced drive from 
die collectors to a single-ended 50-ohm 
termination. Note diat an impedance of 
25 ohms is presented to each collector. 



Fig. 34 — Circuit o( a push-pull broadband amplifier of the 4- to 6-W class. Filtering is 
necessary at the output of this amplifier and the one in Fig. 33 to prevent harmonics from 
being radiated by the antenna system. Q1 and Q2 are GE D44C6 units. Tl, T2 and T3 
contain 6 bifilar turns of No. 26 enamel wire (twisted pairs) on Amidon FT-37-61 toroid 
cores. T4 is the same as Tl, but two cores are stacked. T5 has 6 bifilar turns of a single 
twisted pair of No. 26 enamel wire on two stacked FT-37-61 toroid cores. 


62 Chapter 4 




0- 

le 

is 

ill 

as 

e, 

>y 

is 

re 

er 

is- 

n- 

Jt 

)d 

PS 

.5 

in 

s. 

af 

id 

n- 

'4 

at 

es 

k- 

le 

7 

:h 

ig 


s. 

as 

of 

r- 

le 

>g 

10 

1. 

n 

m 

of 

IT. 



General-purpose broadband push-pull amplifier. This view shows the breadboard version of the 
circuit. The transistor mounting bolts affix the transistors to the heat sink and extend through 
the pc board. Insulating washers are used. The network at the left was used for filtering during 
initial tests on 20 meters. Power output is 6 watts from 1 .8 to 14 MHz. In excess of 20 watts 
can be provided by this amplifier at 7 MHz if a 24-volt power supply is used. 


The push-pull amplifier was tested 
on 40 and 20 meters. At 7 Mil/, the 
measured output was 5-1/2 watts with a 
drive of 0.5-watt. The efficiency was 59 
percent. These measurements were with 
V cc = 12.5 volts. With a 24-volt supply, 
over 12 watts of output were obtained 
with 0.5 watt of drive power. On 20 
meters, 5 watts of output were seen 
a 12.5-volt supply. However, 1 watt of 
drive svas required. While only 7 dB of 
power gain is marginal, it is still useful. 
The amplifier should perform well on 
80 meters, and« delivered 18 W on 160 
while using a 24-V supply. 

Both amplifiers should be followed 
by a filter to remove harmonics. The 
half-wave filters described earlier should 
be adequate. WA7ML11 built one of the 
single-ended amplifiers with half-wave 
filters for 80 and 40 meters. The output 
low -pass filters are selected by means of 
a slide switch. A relay is included to 
switch around the unit during low- 
power operation. 

A Design Exercise 

Assume that a 7-W amplifier is 
needed for 160 or 80 meters. To mini- 
mize the chance for high levels of 
even-order harmonic output a push-pull 
circuit is chosen. Another criterion is to 
design for low cost, particularly with 
respect to the transistors and heat sinks. 
Available driving power is approxi- 
mately 1 to 2 watts. Some measure of 
burn-out protection is wanted should a 
high output SWR occur. Finally, the 
amplifier should cover at least 100 kHz 
of either band without need to retune 
die collector tank. 

The foregoing may seem like a 
tough assignment, especially if under- 
taken by a beginner. Actually, the 
chore is easier than it may seem. 
Nearly all of the information needed 


to effect such a design has been pro- 
vided in the preceding text. Transistor 
selection, network design, and heat 
sinking have been tested in sufficient 
depth to make a simple amplifier 
design possible. 

Transistor Choice 

It was stated earlier that the tran- 
sistors used in an amplifier should carry 
a Pp rating of approximately twice the 


desired rf power output. Therefore, to 
extract 7 W we should use a pair of 
transistors whose combined power- 
dissipation rating will be 14 W or 
greater. Also, the fr should be several 
times the highest operating frequency (5 
or 1 0 times as a ball-park number). This 
calls for an f r of 17 to 35 MHz, or 
thereabouts. Maximum voltage ratings 
should be somewhat greater than two 
times the operating voltage, which sug- 
gests a safe value of 30 or more. 

A search through various data 
showed that an RCA 2N5320 should do 
the job nicely. The price per unit is 
roughly $1.50, f T is 50 MHz, and 
maximum collector voltage is 100. 
Maximum dissipation at 25* C is 10 W 
for a 2N5320, providing a 20-W rating 
for two of the units. The junior version 
of the 2N5320 might be used for a 5-W 
maximum output power in the push- 
pull amplifier of Fig. 35. The device is a 
2N2102, designed specifically for high- 
speed, high-voltage switching. It has an 
fr of 120 MHz, which makes it suitable 
from 1 .8 through 14 MHz. The price tag 
is approximately SI , and the P D (max) is 

Networks 

For the sake of simplicity a conven- 
tional broadband transformer, T1 of 
Fig. 35, is selected for the amplifier 
input port. It will have a turns ratio of 



Fig. 35 - Circuit lor the push-pull 7-W output design example treated in the text. Details for 
a practical heat sink arc shown here. 


Power Amplifiers and Matching Networks 63 







r 


Table 2 


BAND 

LI 

L2 

Cl 

C2 

7 MHz 

0.6 pH, 13 T 

No. 22 enam., 5/16" 

ID, no core 

1.1 pH, 14 T 

No. 22 enam. on 
T-68-2 toroid core 

450-pF 

mica 

trimmer 

820-pF 
silver mica 

14 MHz 

0.3 pH. 8 T 

No. 22 enam., 5/16” 

ID, no core 

0.55 pH. 9 T 

No. 22 enam. on 
T-68-6 toroid core 

450-pF 

mica 

trimmer 

220-pF 
silver mica 

21 MHz 

0.19 pH. 5 T 

No. 20 enam., 5/16" 

ID, no core 

0.39 pH. 6 T 

No. 22 enam. on 
T-68-6 toroid core 

450-pF 

mica 

trimmer 

None 


LI coils are airwound. L2 coils are on Amidon toroid cores. 


approximately 22:1 for a 2 ratio of 
5:1, assuming a total secondary imped- 
ance of 10 ohms (a close approximation 
for a base-to-base impedance of 10 
ohms). The primary inductance should 
be at least 17 /all for 50 ohms at 1.8 
MHz, or 9 fill for 3.5 MHz (*/. of 4 
times 50 ohms = 200 ohms). A 3/8-inch 
diameter Amidon ferrite toroid with a p 
of 125 will be suitable when wound 
with sufficient No. 28 enamel wire to 
obtain the necessary inductance. The 
number of secondary turns is ratio- 
related to the 2.2:1 figure, and are set 
by the number of primary turns neces- 
sary to obtain an X/ of 200 ohms or 
greater. 

A 10-ohm, 1-W resistor is connected 
from each transistor base to ground. 
This will help stabilize the amplifier by, 
lowering the Q of Tl. Final adjustment 
of Tl can be made with the amplifier 
operating at rated output power into a 
50-ohm resistive load. An SWR indica- 
tor is placed between the exciter and 
Tl; then the primary turns of Tl are 
reduced or increased until an SWR of 
approximately 1 is obtained. 

A balanced -collector choke is needed 
for T2. Since the collector-to-collector 
impedance for 7 W of output is roughly 
44 ohms for a 12.5-V dc supply, the 
choke should have an Xi of approxi- 
mately 175 total, or 88 per half. That 
comes to 15 fill at 1.8 MHz, or 7.7 pH 
at 3.5 MHz. The wire size for the 
winding should be able to pass the 
collector dc current without causing an 
/ X R drop. Each transistor will draw 
approximately 0.6 A at tire rated dc- 
input power level, suggesting that No. 
24 enamel wire will be suitable. 

T2 can be wound biftlar fashion (8 


twists per inch of wire) on a piece of 
ferrite rod (Amidon 0.5-inch diameter 
stock) about two inches in length, or on 
a 1 -inch diameter ferrite toroid core. QI 
ferrite will be suitable (p = ’125) in 
either case. The phasing should be as 
shown in Fig. 35. 

T3 is a conventional transformer 
wound with No. 24 enamel wire to have 
a primary inductance of approximately 
the same value used at the primary of 
Tl. The secondary winding of T3 
should have the same inductance. 
Although a calculated Z ratio of 1.13:1 
is appropriate for T3, and 1:1 ratio 
(total primary to secondary) will be 
acceptable. A 3/4- or 1-inch diameter 
QI ferrite core will be adequate at T3. 

LI and L2 can be Amidon 
powdered-iron cores (T-68-2), wound 
with sufficient No. 24 enamel wire to 
provide the required inductance. A 
loaded Q of 4 was chosen for the T net- 
work to assure ample bandwidth and min- 
imum chance for amplifier instability. 


Cl is a large mica compression trim- 
mer of 1000-pF maximum capacitance. 
A J. W. Miller No. 160-A was used in the 
ARRL test model. Fixed-value silver- 
mica capacitors can be used in place of 
the trimmer by combining them to 
ob tain 835 or 417 pF,as specified on the 
diagram. 

RFC1 is a dc decoupling choke of 
low inductance value. A 10-pH value 
will suffice for either band. It can be 
made by winding a 0.5-inch diameter 
QI toroid core full with No. 24 enamel 
wire. 

Zener diodes are used at each col- 
lector for transistor protection in the 
event of a severe mismatch. The diodes 
will have no effect upon performance 
during normal conditions. They need not 
be included if it is unlikely that a high 
SWR will be seen. 

Heat Sinks 

Each transistor will need its own 
heat sink. A simple homemade variety 



Fig. 36 - Schematic diagram of the 15-watt amplifier. Fixed-value capacitors are disk ceramic unless otherwise noted. Resistors are 1 /2-watt 
composition unless specified differently. The 47-pF capacitor can be electrolytic or tantalum. 

Cl — 450-pF mica compression trimmer transistor. 

(Arco-EI-Menco 466 or equivalent). RFC1, RFC2 — 7 turns No. 20enam. wire on 

C2 — See Table 2. 0.5-inch OD toroid ferrite core with 125 

LI, L2 — See Table 2. permeability (Amidon Assoc. FT-50-61 

QI — Motorola MRF449A strip-line stud core or equiv.) , 3 pH. 


Tl — Primary, 32 turns No. 24 enam. on 
Amidon T-68-2 core (7pH). Secondary, 8 
turns No. 24 over primary winding. 




64 Chapter 4 




FOIL SIDE 
(FULL SCALE) 



Fig. 37 — Scale layout of the 15-W amplifier pc board. Doubte-clad board (copper on both sides) 
is used, and the ground foil on the etched side is connected to the ground-plane side at several 
points. Details are given for the homemade heat sink. 


can be fashioned from 2-inch sections of 
Reynolds hardware-type aluminum 
angle bracket (sec sketch in Fig. 35). 
The heat sinks have clearance holes for 
the transistor cases, and a snug fit is 
necessary to assure proper heat transfer. 
Silicone grease should be placed on the 
transistor body where it mates with the 
sink. Each sink is isolated from ground 
by means of insulating-spacer washers. 
The foil on the hottom of the pc board 
should be removed so that the 6-32 
mounting nuts are isolated from ground. 
The foil material on the top of the pc 
board should be removed where the 
6-32 bolts pass through it. The heat 
sinks are snugged down against the 


transistor flanges by means of the 
mounting bolts. 

Results 

A laboratory breadboard of the cir- 
cuit was built and tested for 1 .8 and 3.5 
MHz. Performance was smooth (no in- 
stability), and an output of 8 watts was 
obtained on 160 meters. A 7-W output 
was secured on 80 meters. Examination 
' of the output waveform showed a clean 
sine wave on both bands. Second har- 
monic energy was down some 40 dB, 
and all other harmonics were at least 50 
dB below the fundamental frequency. 
When the amplifier is loaded properly 
into 50 ohms, the tuning of Cl will be 


fairly broad, but a definite peak in 
output will occur when it is set cor- 
rectly. 

15-Watt HF-Band Amplifier 

One advantage of high-gain tran- 
sistors is that they can provide con- 
siderable output power for low-drive 
levels. The Motorola MRF449A is one 
choice a designer has among the high- 
gpin hf-band devices. It is designed for a 
power output of 30 W maximum, Class 
C, when used below 30 MHz. A 13-V 
power supply is required. Power gain is 
rated at 13 dB at 30 MHz. 

The circuit of Fig. 36 shows how it 
can be used in a single-band cw ampli- 
fier with an efficiency of 60 percent. 
The circuit was described originally in 
QST for December, 1975, where it was 
specified as a plug-in amplifier for the 
Heath HW-7 QRP transceiver. The 3-dB 
resistive attenuator at the amplifier 
input is included so that exciters having 
more than 1 watt of output will not 
overdrive the transistor. The HW-7 de- 
livers 2 watts of output, so 1 watt is 
absorbed in the attenuator. Also, the 
attenuator provides a constant 50-ohm 
load for the exciter. The addition was 
necessary because the MRF449A re- 
quires only 3/4 to 1 watt of drive to 
produce full output. Those having ex- 
citers in the 1 -watt class can delete the 
attenuator. 

T1 is a conventional input trans- 
former which is wound on a T-68-2 
powdered-iron toroid. It provides a 
necessary 16:1 transformation ratio (50 
to 3 ohms). Two 4:1 broadband trans- 
mission-line transformers were tried in 
cascade to replace Tl, and results were 
identical to those with the transformer 
specified. The conventional transformer 
was used because only one. toroid was 
required. To lower the Q of Tl a pair of 
10-ohm, 1/2 -W resistors have been 
strapped from base to ground. 

Power Level 

A power-output level of 15 watts 
was chosen to minimize power-supply 



Fig. 38 - Photograph of the assembled 
amplifier. The circuit-board pads of Fig. 37 
replace the phono plugs shown here. 


Power Amplifiers and Matching Networks 65 





Fig. 39 - Schematic diagram of the 15-watt linear amplifier. Resistors are 1/2 -W composition unless otherwise noted. Capacitors are disk 
ceramic unless specified differently. Polarized capacitors are electrolytic or tantalum. T1 

RFC1 — Miniature 1.5-pH choke. 2:1 impedance ratio. 14 turns No. 28 No. 24 enam. wire (bifilar wound to 8 

8FC2 — 15 turns No. 26 enam. wire on enam. wire on Amidon FT-50-61 toroids turns per inch) for winding 1/3/4. Winding 

Amidon FT-37-61 toroid. (two cores stacked). Secondary has 10 2/5 contains 15 turns of single No. 24 

RFC3 - 7 turns No. 20 enam. wire on turns of No. 28 enam. wire over primary enam. wire. Use two Amidon FT-50-61 

Amidon FT-50-61 toroid. winding. cores, stacked. 

T1 - Conventional broadband transformer, T2 — Broadband 1:1 transformer. 15turns 


drain for field use. The network values why the builder could not develop band transformers of Fig. 39 were devel- 
are based on that power amount (Table suitable L and C values for the T oped. Performance remains essentially 
2), but there is no reason why the full network from the reactances listed in the same regardless of the transformer 
30-W output amount cannot be realized. Fig. 25. At 80 and 160 meters there style employed. Lab tests with a spec- 
The collector network would have to be may be a tendency toward instability, trum analyzer show that both versions 
revised accordingly. If that were done, a owing to the higher gain of the tran- provide an IMD (3rd- and 5th-order 
collector characteristic of 2.8 ohms sistor at those frequencies. An addi- products) of - 30 dB. 
would result. Therefore, a T network tional 10-ohm resistor from base to A peak output power of 15 watts is 
with a loaded Q of 4 would require an ground should resolve the problem, available on ssb, and 1 5 watts of output 
XL\ of 1 1 . an XL2 of 7, and an XC\ of Alternatively, the negative-feedback are provided lor cw work. Forward bias 
12. The circuit was tested at the 30-W technique shown in Fig. 25 could be is supplied to the transistor bases to pre- 
level and performance was good, applied to enhance stability. vent cross-over distortion. (See chapter 

However, a slightly larger heat sink than The output waveform as viewed on a 8.) Idling current (no drive applied) is 
that shown in Fig. 37 will be necessary 50-MHz scope was very clean. Harmonic approximately 100 mA with 28 volts of 
at the higher power amount. The dimen- energy was at least 40 dB below carrier collector supply. Peak current drain is 
sions for Tl, RFC1, and RFC2 are level. Fig. 38 is a photographic view of 1.5 A. 

suitable for either power value. A 50-W the module. Although the amplifier is designed 

version of the '449A is available for for 3.5 to 30 MHz. good performance 

those wanting more power. It is the A 1 5-Watt Linear Amplifier was no t e( i on 1 60 meters with approx- 

MRF450A. Approximately 2 watts of The amplifier of Fig. 39 was adapted imately 1 watt of drive. The original 
drive power are needed for full output, from one which was described by Lowe version by Lowe was not tested at 1.8 
Operating voltage is 13 for the latter (jQST for Dec., 1971, p. 11). The basic MHz, however. 

also. Both transistors are stud-mount difference is in the transformer design The input port contains a complex 
types, and each has strip-line connecting (Tl and T2). The Lowe transformers RCA compensating network to level the 
leads. were similar to that of Fig. 28 in this amplifier gain by compensating the 

Specifications are not given for 160, chapter, but many amateurs had dif- drive level. Amplifier gain is 16 dB at 15 

80, or 10 meters, but there is no reason Acuity duplicating them, so the broad- and 10 meters, and is slightly greater on 


66 Chapter 4 














MC1723G 


• R2 
1000 
BIAS 
AD J 


Fig. 40 - Schematic diagram of the 300-W output linear amplifier designed by Granberg. 
Capacitors are the ceramic chip variety except for Cl 1 , which is electrolytic. Numbered 
components not described here are so identified for layout purposes on the pc-board pattern 
offered in the QST series. 


LI. L2 — Rf choke (Ferroxcube VK200-19/ 
4B or equiv.). 

L3, L4 — Rf choke (Ferroxcube 56-590-65/ 
3B or equiv.). For these chokes and other 
Ferroxcube components contact Elna 
Ferrite Labs., Inc.. 9 Pine Grove St., 
Woodstock. NY 12498. 

T1 - Broadband 9:1 transformer on ferrite 
core (TV balun type Stackpole 57-1845- 
24B. Fdir-Rite 2973000201 , or Amidon 
equivalent of latter). Low-^ winding has 
one turn of 1 /8-inch OD copper braid to 
serve as tubing. Primary contains three 


turns of No. 22 Teflon or enamel-coated 
wire. 

T2 — 7 bifilar turns of No. 22 enam. wire on 
Stackpole 57-9322 or Indiana General 
F627-8Q1 toroid core. A suitable substi- 
tute core would be two Amidon FT-50-61 
cores, stacked. 

T3 - 14 turns Microdot 260-4118-000 25-ohm 
submin. coaxial line or equiv., wound on 
each of two Stackpole 57-9074 or Indiana 
General F624-19Q1 cores. A probable sub- 
stitute is the Amidon FT-1 14-61 toroid. 




20, 40 and 80 meters. Input SWR 
through the compensating network is 
less than 1.5:1 from 80 through 10 
meters. 

Q1 and Q2 are low-cost surplus vhf 
transistors. The 2N3632 is designed for 
Class A, B and C service. Maximum 
Vceo > s 65, maximum collector current 
is 3 A, and fr is 400 MHz. Maximum 
dissipation is 23 W at 25°C. 

A finned heat sink measuring 4 X 4 
inches or greater is required for safe 
operation. Double-sided pc board is 
used to contain the amplifier. Output 
SWR should never exceed 1.5:1 if 
damage to the transistors is to be 
prevented. Although the even-order har- 
monics from the amplifier are at least 
20 dB below the fundamental signal, 
filtering should be used at the output. 
The half-wave filters described earlier in 
the chapter will be suitable. 

A 300-Watt-Output Linear 
Amplifier 

This chapter would not be complete 
without an example of a high-power 
linear amplifier. The circuit of Fig. 40 
shows a design by H. Granberg 
(WB2BHX/7) of Motorola, which is one 
module of a 1200-W composite ampli- 
fier (four power blocks combined). He 
described the latter in QST for April 
and May, 1976. 

This circuit contains two Motorola 
MRF428A transistors. An operating 
voltage of 50 is required and current 
taken is approximately 13 A. The cir- 
cuit is broadbanded for use from 1 .8 to 
30 MHz. Full output can be obtained 
with a driving power of 5 W, as observed 
in ARRL laboratory tests. Harmonic 
filtering is required at the amplifier 
output during on-the-air use. 

The module contains a bias regulator 


TEMP __ 
SENSE 


Photograph of the assembled 300-W amplifier. Note 1/4-inch thick copper plate between the 
double-sided pc board and the aluminum heat sink. 


Fig. 41 — Schematic diagram of the bias regu 
lator and temperature sensor. 


Power Amplifiers and Matching Networks 6 









to provide forward base voltage for 
linear operation. Fig. 41 shows the 
circuit. Variable bias voltage is available 
by means of R2, providing a range from 
0.5 to 1 V, regulated. CR1 is the 
base-emitter junction of a 2N5190. It 
has a plastic case and is used as a 
circuit-board standoff spacer. It serves 
as a temperature-sensing diode. By 
virtue of its being coupled to the heat 
sink it assures automatic temperature 
tracking with a slight negative coef- 
ficient. When the collector idling cur- 
rent is set for 300 mA at 25°C, the 
current will decrease to a nominal 250 
mA when the sink temperature rises to 
60°C. The rate of change is approxi- 
mately 1.15 to 1.7 mA per degree C. 

In Fig. 40 a 9:1 input transformer is 
used, providing an impedance step-down 
from 50 ohms to a 5.5-ohm base-to-base 
characteristic. Negative feedback is em- 


ployed to enhance stability and to help 
equalize amplifier gain. Approximately 
5 to 6 dB of feedback can be utilized 
without impairment of linearity or sta- 
bility. 

In addition to providing a source for 
negative feedback, T2 supplies dc volt- 
age to the collectors and serves as a 
center tap for output transformer T3. 
The currents for each half cycle are of 
opposite phase in ac and bd, and de- 
pending upon the coupling factor be- 
tween the windings, the even-harmonic 
components will see a much lower 
impedance than will the fundamental 
energy. The resonant frequency of 
C5-L5 should be above the highest 
operating frequency to prevent insta- 
bility. 

A 4:1 transmission-line transformer 
is used at T3. It is a coaxial-cable type, 
with a and b wound on separate cores. 


This technique eliminates the need for 
having three separate transmission lines, 
which would be the requirement if a 
single core were used. The line sections 
consist of 25-ohm miniature coaxial 
cable with Teflon insulation. Alterna- 
tively, twisted pairs of enamel-coated 
wire can be used to form 2 5 -ohm lines 
(discussed earlier in this chapter), but 
the coaxial cable specified is recom- 
mended strongly by Granberg. 

Heat sinking is of extreme impor- 
tance in this amplifier. The transistors 
are joined thermally to a thick block of 
copper plate, and the latter is coupled 
to a large aluminum heat sink. Chip 
capacitors are used throughout the rf 
circuit of the amplifier. Most of them 
are on the bottom side of the pc board. 
The reader is referred to the original 
QST material if duplication of this 
circuit is anticipated. 


68 Chapter 4 


Chapter 5 


Receiver Design Basics 


The most used piece of equipment in 
any amateur station is the receiver. 
During communications with other sta- 
tions the receiver accepts signals from 
the antenna to produce intelligible 
audio output. At other times, the re- 
ceiver is used to “scan the band" and 
monitor QSOs. The station receiver is 
also a valuable piece of test gear. 

In the early days of amateur radio, it 
was necessary for every ham to build his 
own receiver. However, by the time the 
1930s arrived, it was common to find an 
amateur station with homemade trans- 
mitting equipment and a commercially 
built receiver. This was the rule rather 
than the exception in the early 1950s 
when the writers first became licensed. 
Tie onslaught of single sideband prior 
to the ’60s brought with it the “ap- 
pliance era.” when few amateurs built 
their transmitters, much less their re- 
ceivers. The complexity of each was 
similar, making home construction a 
task for only the more ambitious and 
enthusiastic. 

The dominance of semiconductor 
technology has changed this. Today it is 
straightforward to build receivers of 
simple design while using transistors and 
lCs. Even receivers offering something 
approaching state-of-the-art per- 
formance are constructed easily if the 
builder is willing to invest in a bit of 
time and experimentation. 

In spite of the relative ease of 
construction, some amateurs are not 
willing to build a receiver. This is 
unfortunate, for one of the most ex- 
citing experiences available to the ham 
is the thrill that results from using a 
receiver he has constructed himself. 

In this chapter we will discuss some 
basic ideas associated with cw and ssb 
receivers. For the most part, the empha- 
sis will be on straightforward and simple 


approaches to design. Several practical 
examples are presented. 

In chapter 6 we will consider some 
refined details of receiver design. The 
emphasis will be on designing for wide 
dynamic range. The reader is referred to 
the transceiver section of the book for 
additional construction information. 

Fundamental Considerations 

Certain criteria must be met in tire 
design of a receiver of even the simplest 
kind. These include meeting specifica- 
tions for gain, selectivity, sensitivity and 
stability, to mention only a few. 

The first requirement for a receiver 
is to provide considerable gain. Tire 
signal levels from the antenna are often 
quite low, while enough power output 
to drive a speaker or a pair of head- 
phones is ultimately desired. If we 
assume a weak cw signal as being 0.1 pV 
available from a 50-ohm antenna, the 
power available to the receiver is 

p- H = n x i o ~ 7 ) 2 
R 50 

= 2 X 10" 16 watts (Eq. 1) 

If we would like this signal to produce an 
output of 1 volt across a pair of 2000-ohm 
headphones, the output power is 5 x HH 
watts, or half a milliwatt. The neces- 
sary power gain is then the ratio of these 
powers 

c= LXJ0r* 2.5X10 12 (Eq-2) 

2 X 10'“ 

This is 124 dB and is typical of the net 
gain in many receivers. Since the signals 
of well under the 1-volt output men- 
tioned above are copied easily in 2-kf2 
headphones, less gain is often satis- 
factory. Around 80 to 90 dB is prob- 


ably an absolute minimum for com- 
munications applications. 

A second requirement for a receiver 
is that it process the incoming signal to 
cause an audio voltage to appear at the 
output. The process is called detection. 
Circuits to perform this function will 
vary considerably, depending upon the 
nature of the information contained on 
the incoming signal. 

In all of the receivers described in 
this chapter, product detection is em- 
ployed. A product detector is really 
nothing other than a mixer (chapter 3). 
However, the two signals to be mixed 
are that of a beat-frequency oscillator 
(BFO) and a second signal closely 
spaced. Tire output of the mixer is at 
audio frequencies. The term “product 
detection" results from the character- 
istic of a mixer that the amplitude of 
the output signal is proportional to the 
product of the two incoming signal 
voltages. In most situations, the BFO 
level is very much higher than the 
incoming signals to be detected, often 
by 100 dB or more. Under these con- 
ditions, the detector is essentially a 
linear device in that the output of the 
detector is directly proportional to the 
amplitude of the input. This is not the 
case for a-m detectors where a threshold 
exists, or for fni detectors where the 
output is independent of incoming am- 
plitude once a suitable threshold is 
overcome. The linearity of a product 
detector is of profound significance, for 
it allows us to achieve tremendous 
simplification in designing simple cw 
and ssb receivers. 

Another characteristic which a re- 
ceiver must possess is selectivity. That 
is. it must be capable of isolating two 
signals which are closely spaced in fre- 
quency. This is realized with filters of 
various kinds, either at radio or audio 


Receiver Design Basics 69 


r 


frequencies. Both filter types are dis- 
cussed later in this chapter. 

Along with selectivity, a receiver 
must exhibit stability. The stability re- 
quired will depend primarily upon the 
selectivity of the receiver, with the 
gpneral criterion that the drift in the 
tuning should be small in comparison 
with the bandwidth of the receiver. The 
problems of long-term oscillator sta- 
bility were outlined in the discussion of 
VFOs in chapter 3. 

Another receiver parameter is sensi- 
tivity. This is usually specified by noting 
the signal power (available at the input 
to the receiver) required to yield a given 
output signal-to-noise ratio. The gain 
calculations outlined earlier might imply 
that the sensitivity of a receiver can be 
made arbitrarily low by providing more 
and more gain. Such is not the case. The 
culprit, in this case, is noise. Any 
amplifying device will have some noise 
generated in it. This noise will add to 
the signals in the output to cause a 
degradation in the output signal-to-noise 
ratio. 

A measure of the degradation of 
signal-to-noise ratio caused by an am- 
plifier or receiver is the noise figure or 
noise factor. The formal definition of 
the noise factor of an amplifier is given 
as 


NF=^ (Eq.3) 

^out 

Nout 

where the input and output signals and 
noises are powers in watts. If the ratio is 
calculated as shown above, the term is 
usually called noise factor. If the power 
ratio is. however, expressed in dB. the 
term noise figure applies. 

The output signal and noise powers 
are, in principle, easily measured. Simi- 
larly, the input signal power available 
from a quality signal generator is well 
defined. Hcwever, the input noise 
power is not as well defined. As a 
standard, the input noise power is 
usually assumed to be the power avail- 
able at the terminals of a resistor at a 
temperature of 290 degrees Kelvin. The 
power, P„, is given by 

P„=kTB (Eq.4) 

where T is the temperature in degrees 
Kelvin, B is the bandwidth in Hz and k 
is Boltzman’s constant, 138 X 10' 23 
watts/deg.-Hz. 

Consider a simple receiver, as an 
example, to illustrate the noise-figure 
concept. Assume that the gain of the 
receiver is 100 dB and that the band- 
width is 500 Hz. If a 50-ohm resistor is 
attached to tire receiver, the noise 

70 Chapter 5 


power available at tire antenna terminal 
iskTB = 1.38 X 10' 23 X 290 X 500 = 2 
X 10' 18 watts. If this receiver were 
perfect, with no internally generated 
noise, the output noise power would be 
10 1 " (100 dB) times this value, or 2 X 
10' 8 watts. However, the receiver, 
being a real system, does generate some 
noise of its own. Hence, the output 
noise pcwer will be somewhat higher. 
Assume that the output noise is 1 X 
10' 7 watts. 

If we note the equation for noise 
factor, we see that it may be rewritten 
as a ratio of “noise gain” divided by 
“signal gain.” 

NF = % (Eq. 5) 

Substituting the above noise powers 
in for the noise gain, that is, the noise 
output divided by the noise input, we 
see that G„ = 5 X 1 0‘ 0 while the signal 
gain was only I X 1 0 1 0 , or 100 dB. 
Hence, the noise factor is 5: The noise 
figure is merely 10 log (noise factor), 
or 7 dB. This value is quite typical for 
the better communications receivers 
operating in the 3- to 30-MHz region. 

The foregoing arithmetic can be 
worked backward to tell us what the 
minimum signal level is that may be 
detected with this receiver. The noise 
output of the receiver was 10 ~ 7 watts 
and the gain was 100 dB. Hence, a signal 
at the input which was 100 dB below 
10~ 7 watt, or 10"' 7 watts would yield a 
unity output signal-to-noise ratio. This 
signal corresponds to about .02 micro- 
volt across a 50-ohm resistor. A signal of 
about 02 microvolt would yield a 20 
dB signal-to-noise ratio at the output. 

There are a number of factors to be 
learned from this analysis. First, the 
lower the noise figure, the more sensitive 
the receiver will be. Of equal signifi- 
cance, the narrower the bandwidth, the 
less noise will get through the receiver 
and the more sensitive it will be. How- 
ever, the bandwidth of a receiver can be 
decreased only to the point where it is 
the same as the bandwidth of the 
information to be recovered by the 
receiver. This explains why cw is so 
much more effective during weak-signal 
conditions than is any form of phone, 
including ssb. 

There is another factor which does 
not drop immediately from this anal- 
ysis. Often, with experienced and capa- 
ble cw operators, it is found that signals 
can be copied which are much lower 
than a measurement of receiver sensi- 
tivity would suggest being possible. This 
is demonstrated easily with a good 
receiver with switchable bandwidths and 
a signal generator. The receiver is first 
set at the narrowest bandwidth available 


and the signal generator is adjusted to 
deliver the weakest possible cw signal 
which the operator can perceive. Then, 
the bandwidth of the receiver is in- 
creased to the widest available setting. 
More often than not, the operator can 
still hear the signal. The reason for this 
apparent discrepancy is that the oper- 
ator, or listener, is part of the receiving 
system. His mental process essentially 
forms a very narrow bandwidth, adap- 
tive (i.e., learning) filter. 

This rather subtle effect is not 
merely a curiosity of nature. It can be 
used effectively to copy amazingly weak 
signals from simple receivers. Alterna- 
tively, it can be used for the copy of 
extremely weak signals which might 
never yield usable output on a meter, 
The most profound examples of this are 
the day-to-day moonbounce contacts 
which arc made by means of advanced 
vhf and uhf amateur stations. The re- 
ceivers used at such stations have band- 
widths of 2 kHz down to perhaps 100 
Hz, and exhibit noise figures of 1 to 2 
dB. 

Rarely on the hf bands is a low noise 
figure needed in a receiver. The reason 
for this is that the man-made and 
atmospheric noise levels found in most 
locations are so high that they mask any 
noise generated within the receiver. This 
factor can be used to advantage by the 
experimenter. It doesn't matter what 
the ultimate numerical value for receiver 
sensitivity is. There is one experiment 
which is more significant: Disconnect 
the antenna of the receiver and listen to 
the noise output of the receiver. Then, 
connect the antenna and listen to the 
background noise. If the noise increases 
dramatically, the sensitivity of the re- 
ceiver is as good as it needs to be. That's 
all that counts! (Strictly speaking, llie 
antenna should be replaced with a 50- 
ohm resistor for comparison, although 
this is rarely of importance with hf 
receivers.) 

Even though low-noise-figure re- 
ceivers are rarely needed for the hf 
bands, the concept is quite important in 
the design of high-performance re- 
ceivers. This is especially true if it is 
desired to design a wide-dynamic-range 
receiver. An overview of the noise-figure 
concept has been presented here, 
Further information is given in chapter 
6. 

Block Diagrams 

There are essentially two forms 
which the block diagram of an hf 
receiver can take. They are the classic 
superheterodyne and the direct- 
conversion receiver or synchrodyne, 
Shown in Fig. 1 is a block diagram for 
the latter, a design which has been 
popular in this country since 1968. The 
signals from the antenna are applied to 
the input of the receiver through a 


T 



Fig. 1 - Block diagram of a direct-conversion receiver. 


simple bandpass filter. The output of 
this filter is routed to a product detec- 
tor which is driven by a BFO voltage 
which is very near the frequency to be 
received. The output of the detector is 
applied to a low-pass filter, then routed 
to a high-gain audio amplifier, thus 
completing the receiver. The advantage 
of this approach to receiver design is the 
extreme simplicity afforded. The 
number of stages is minimized. Most of 
the gain is obtained at audio fre- 
quencies, where construction is simple. 
Finally, the BFO operating at virtually 
the same frequency as that of the 
received signal leads to the design of 
simple transceivers. 

There are other advantages to the 
direct-conversion concept which will be 
described later. However, there is a price 
to pay for all of this simplicity — the 
receiver is not a panacea. Consider, as an 
example, a signal to be received at 7049 
kHz. The BFO might be set to 7050 
kHz, resulting in a 1-kHz beat note from 
the detector. This signal is amplified in 
the audio stages of the receiver and 
applied to the headphones. 

Consider the response to signals at 
other frequencies. For example, a signal 
at 7040 kHz would not be attenuated 
by the front-end bandpass filter. Hence, 
it would also be applied to the input to 
the product detector and would result 
in an output beat note of 10 kHz (the 
BFO is still at 7050). The low -pass filter 
will prevent most of the 1 0-kHz energy 
from arriving at the audio amplifier, so 
this signal causes no significant problem. 
Consider now, a signal at 7051 kHz. 
This signal will reach the input of the 
detector and heterodyne with the BFO 
output at 7050 kHz to produce a 1 -kHz 
beat note, which is exactly the same 


response as obtained from the desired 
signal at 7049 kHz. Hence, no amount 
of audio filtering will eliminate this 
response. This undesired response is 
called an audio image, and it is a major 
disadvantage with direct-conversion de- 
signs. In spite of this, thousands of 
amateurs have built “dc” receivers and 
use them daily. The simplicity of design 
is worth the few practical problems 
which arise from the audio image during 
routine communications. Although the 
existence of the image would have the 
effect of doubling the equivalent noise 
bandwidth of the receiver, this effect is 
largely negated by the filtering nature of 
the human ear. There is virtually no 
fundamental sensitivity penalty to be 
paid for the use of direct-conversion 
receivers. 

Shown in Fig. 2 is a block diagram 
for a classic superhet receiver. Here, the 
incoming signal is applied to a pre- 
selector bandpass filter and is then 
routed to a mixer. The mixer is also 
driven by a local oscillator which is 
separated from the incoming frequency. 
The output of the mixei is at a fre- 
quency which is tire difference (or the 
sum) of the incoming signal and the 
local oscillator (LO). This frequency is 
called the intermediate frequency, or i-f. 
The i-f output from the mixer is applied 
to a filter which usually has a band- 
width compatible with the signals being 
received. The i-f signal is amplified 
further before it is applied to a product 
detector. The detector output is ampli- 
fied and then applied to headphones or 
a speaker where the user should perceive 
some intelligent information. 

Consider a receiver with an i-f of 1 
MHz. Assume that the i-f filter has a 
bandwidth of 500 Hz and suppose that 



Fig, 2 - Configuration for a basic superheterodyne receiver. 



this receiver is tuned to the same signal 
at 7049 kHz that was used in the “dc” 
receiver example. For the 7049-kHz 
signal to be received, the LO will be 
tuned to 6049 kHz, resulting in a 
1000-kHz output i-f signal. This signal 
moves readily through the 500-Hz-wide 
filter, is amplified and detected. If the 
detector is driven by a BFO at 999 kHz, 
a 1 -kHz receiver output will result. 

Now consider that same bothersome 
signal at 7051 kHz. This signal will beat 
in the mixer with the local-oscillator 
energy at 6049 kHz to produce an i-f 
output at 1002 kHz. However, the i-f 
filter is only 500 Hz wide. Hence, the 
filter will have significant attenuation at 
1002 kHz, and no receiver output will 
result. The superhet has eliminated the 
troublesome audio image which plagued 
the dc receiver. This asset of a superhet 
is called single-signal response. 

Image responses will still be present, 
but now they are associated with the 
intermediate frequency rather than with 
audio. For example, our receiver has a 
1-MHz i-f and an LO at 6 MHz, for a 
desired input near 7 MHz. However, 
signals at 5 MHz will also beat with tire 
LO to produce 1-MHz i-f signals. Hence, 
everything possible should be done to 
prevent 5-MHz signals from reaching the 
mixer input. This is easily realized with 
the 7-MHz preselector filter. 

The following sections will consider 
design details of the various sections of 
direct conversion and superheterodyne 
receivers. Examples are presented for 
duplication. Emphasis will be on simple 
designs. 

Product Detectors 

The product detector is the basis of 
the direct-conversion receiver, and it is 
an integral part of a “superhet” receiver 
designed for cw or ssb reception. As 
mentioned earlier, a product detector is 
essentially a mixer. As such, it is a 
three-port circuit with two radio- 
frequency inputs and an intermediate- 
frequency output. When a mixer is used 
as a product detector, the i-f is at audio. 
A product detector is shown in block- 
diagram form in Fig. 3. 

When used as the front end of a 
direct-conversion receiver, a product 
detector has a number of necessary 
specifications. First, it must have a 



Fig. 3 — Representation of a product detector. 


Receiver Design Basics 71 









Fig. 4 - CA3028A product detector. 



(200 



Fig. 6 - Application of an MC1496G 1C as a product detector. 


fairly low noise figure (low noise at rf 
and audio frequencies). Some gain is 
sometimes desired, although certainly 
not necessary. The detector should also 
have the ability to handle a wide range 
of signal-input levels without the unde- 
sirable effects of intermodulation dis- 
tortion, blocking and cross modulation. 
Finally, there should be essentially no 
audio output except that which results 
from mixingwith the BFO. 

When used as a detector in a super- 
het, the circuit requirements are some- 
what relaxed. Noise figure is no longer 
of major concern, since the detector is 
usually preceded by circuits with con- 
siderable gain. Often the dynamic-range 
requirements can be relaxed since the 
detector is protected by an automatic 
gain-control (age) system. However, 
intermodulation distortion is still of 
concern, since two signals within the 
passband of the i-f amplifier can pro- 
duce spurious outputs. 

There are a number of circuits which 
offer satisfactory performance as pro- 
duct detectors. It is difficult to say 
categorically which of these is best, for 
all have assets as well as problems. A 
variety of circuits is presented for the 
experimenter to consider. 

Shown in Fig. 4 is a detector popu- 
larized in 1969. It uses an RCA 
CA3028A differential amplifier IC. 
Other similar “pills” could be used, 
These include differential amplifiers 
such as the Motorola MFC8030, and 
transistor arrays such as the RCA 
CA3046. The CA3028 detector is per- 
haps one of the easiest circuits to use, 
since it has a reasonable noise figure and 
considerable gain. For example, direct- 
conversion receivers have been described 
using such a detector, followed by a 
single transistor or IC as the total audio 
amplifier. If maximum gain is to be 
realized with this circuit, the output 
should be terminated in a fairly high 
impedance. This is usually realized with 
an audio transformer with a 10-kfi 
primary. 

Several volts of BFO injection are 
often used with this circuit, resulting in 
a switching type of current waveform at 
the collector of the common current- 
source transistor of the 1C. To optimize 
performance, it is advisable to bypass 
the emitter (pin 4) of this transistor. 

If large-signal problems are en- 
countered with this detector, such as 
blocking or cross modulation, the 
signal-handling properties may be im- 
proved by decreasing the output col- 
lector termination impedance and by 
“standing" additional current in the IC. 
The quiescent current may be increased 
by adding a 330-ohm resistor from pin 4 
of the CA3028A to ground. The output 
termination impedance can be lowered 
by changing the transformer ratio, or by 
using low-value collector resistors in 


Fi< 
Sf 
re I 
me 
nu 


pai 

Th 

eve 

bal 

sigi 

the 

enc 

the 

teri 

In 

per 

ten 

in t« 

pro 

con 

MO 

sen 

use< 

out 

que 

Will 

inje 

mat 

can 

load 

the 

SU P [ 

imn 

mo 

freq 

ofte 

fron 

grad 

con 1 

fide 

with 

stan 

opei 

sign; 

Eure 

USA 

intei 

pres< 

dete 


72 Chapter 5 







Fig 7— A product detector can be built from an 
SN76514 1C. The SM-76514 mixer 1C has been 
reidentified as TL-442-CM by Texas Instru- 
ments. It may be procured under either part 
number. 

parallel with the transformer primary. 
The decreased collector load will, how- 
ever, decrease the detector gain. 

The CA3028A, as shown, is a singly 
balanced product detector. The input 
signals are applied differentially, while 
the BFO drive is applied in a single- 
ended fashion. This tends to minimize 
the BFO energy present at the antenna 
terminals of a direct-conversion receiver, 
in one case where measurements were 
performed, the power at the antenna 
terminal was —47 dBm (.02 microwatt 
into a 50-ohm load). 

Another popular and easily applied 
product detector for use in direct- 
conversion receivers is a dual-gate 
MOSFET (Fig. 5). The circuit is es- 
sentially the same as the mixer circuits 
used with this device, except that the 
output is designed for audio fre- 
quencies, with rf signals being bypassed. 
With detectors of this kind, the BFO 
injection at gate 2 should be approxi- 
mately 5 volts pk-pk. Additional gain 
can be realized by increasing the output 
load impedance. This, however, requires 
the use of transformer coupling or a 
supply voltage well above 12. 

The dual-gate MOSFET has good 
immunity to blocking, 1MD and cross 
modulation. However, the audio- 
frequency noise Figure of MOSFETs is 
often not as good as those expected 
from diodes or JFETs, yielding a de- 
graded receiver sensitivity in direct- 
conversion applications. The major de- 
ficiency of this detector is its behavior 
with a-m signals. The MOSFET is sub- 
stantially a square-law device and will 
operate as a square-law detector of a-m 
signals. This causes severe problems in 
Europe and on the East Coast of the 
USA on 7 MHz, where large signals from 
international broadcast stations are 
present. Proper use of balance in the 
detector should minimize this problem. 


Several ICs other than simple dif- 
ferential amplifiers function well as pro- 
duct detectors. Notable examples are 
the Motorola MC1496G and the Texas 
Instruments SN-76514. The reader is 
referred to chapter 3, where these de- 
vices were applied as transmitting 
mixers. The MCI 496 is used as a pro- 
duct detector by merely replacing the rf 
collector load with a pair of 2.7-kft 
resistors to pins 6 and 9. A circuit is 
shown in Fig. 6. Audio is extracted 
from one of the output terminals through 
a 1 0 -/lt F capacitor. Each of the output 
pins should be bypassed for rf via a 
.05 -/jF capacitor. Additional conversion 
gain can be had by using a center tapped 
transformer at the output. 

The T1 SN-76514 has built-in 600- 
ohm collector resistors. Hence, this chip 
is used as a detector by bypassing the 
two output pins (3 and 13) for rf, and 
by taking audio from one of the pins 
through an electrolytic capacitor (see 
Fig. 7). The relatively low collector load 
resistors in the T1 balanced-modulator 
IC will limit the conversion gain to 
roughly 14 dB, while much more gain 
can be realized from the MCI 496. 

If the internal circuit of the MC1496 
is studied, it can be seen that the input 
signal is injected differentially to a pair 
of transistors with externally applied 
emitter degeneration. The level of this 
negative feedback is controlled by the 
value of resistance between pins 2 and 
3. In the interest of signal-handling 
capability, this resistor should be as high 


in ohmic value as reasonable, perhaps as 
much as 1 000 ohms. On the other hand, 
the resistance should be zero if maxi- 
mum conversion gain and optimum 
noise figure are desired. Hence, the 
value will probably be much different 
for applications in superhets than it 
would be for use as the input to a 
direct-conversion receiver. 

The Motorola applications literature 
of the ’1496 shows the chip biased so 
that about 1 mA flows in each of the 
collector output pairs. However, the 
signal-handling properties of the chip 
can be improved significantly by in- 
creasing the current to approximately 3 
mA in each collector. This is effected by 
changing the usual 10-kH resistor be- 
tween the 12-volt supply and pin 5 to a 
3.3-kd unit. This biasing scheme is 
useful also when the chip is employed as 
the mixer in a ssb transmitter, where 
linearity is of major importance. 

Another IC which functions well as a 
product detector is the RCA CA3102E. 
This IC is a dual differential amplifier 
and is wired externally much like the 
MCI 496 detectors discussed above. A 
circuit is shown in Fig. 8. Good noise 
figure (as well as fine signal-handling 
ability) was observed with this circuit. 
These traits probably result from a lack 
of feedback in the signal input, and the 
switching nature of the circuit, re- 
spectively. The detector circuit shown is 
a doubly balanced format, requiring 
push-pull inputs at the signal and BFO 
ports. A single-ended BFO is converted 



Fig. 8 — A CA3102E can be used as a doubly balanced detector as shown here. 


J 


Receiver Design Basics 73 









to a balanced drive with a ferrite trans- 
former much like those used for bal- 
anced frequency multipliers. The signal 
input is by means of a bifilar link 
around an input tuned circuit. 

Diode Detectors 

There is a class of product detectors 
which have not been described in this 
book. All use diodes as the nonlinear 
element. The experimenter may view 
diode detectors as being useful only in 
special cases where simplicity or a min- 
imum parts count are special criteria, 
thinking, “Such detectors are obviously 
inferior to those using FETs or ICs." 
Nothing could be farther from the 
truth! Detectors (and mixers) using 
diodes are among the best available if 
they are constructed and used properly, 
with good transformers and adequate 
BFO(or LO) injection. 

Shown in Fig. 9 are three detectors 
which use diodes. These circuits contain 
broadband transformers made with tri- 
filar windings on a ferrite toroid core. 
(The reader is referenced to chapter 4 
for details on the construction of this 
type of transformer.) 

The simplest of these detectors is 
that of Fig. 9A. This is a singly balanced 
circuit with the BFO applied at point C. 
Note that a signal at C drives the two 
secondary windings of the transformer 
in opposite directions. Hence, no mag- 
netic field is established in the core. As 
point C swings positive, the upper diode 
is driven into conduction, placing a 
charge on the 0. 1 -/l/F capacitor. But, on 
negative swings of the BFO, the lower 
diode conducts, and a similar charge is 
removed from the capacitor. The overall 
result is that the average voltage across 
the capacitor is zero. However, when a 
signal appears across the transformer, 
one diode goes into conduction slightly 
sooner (or later) than it would have 
otherwise, causing an unbalance in the 
net current flowing into the capacitor. 
Over a period of time, this net transfer 
of charge is observed as an audio voltage 
at the output. The diodes are assumed 
to be virtually identical. 

In the detector at Fig. 9B, two 
diodes have been added. These diodes 
have the effect of presenting a more 
symmetrical load to the BFO, resulting 
in sligluly improved balance and better 
isolation of the BFO from the signal 
circuit. The circuit is still singly bal- 
anced. The circuit shown in Fig. 9C is 
doubly balanced, resulting in good iso- 
lation between all three ports of the 
mixer. 

Detector balance is of minimal sig- 
nificance when the detector is at the 
front of a direct-conversion receiver. 
However, balance can be of considerable 
consequence when used at the detector 
in an advanced superhet. In such a 
design, it is mandatory that the energy 

74 Chapter 5 


from the BFO be confined to the 
detector, and not be allowed to find its 
way into earlier parts of the circuit. If 
extraneous BFO energy gets into pre- 
ceding i-f amplifiers, noise modulation 
may occur, which has the effect of 
creating a “mushy” sounding output 
from the receiver. Having no i-f stage 
preceding the detector in a direct- 
conversion receiver will lead to an ex- 
ceptional signal “cleanliness” and 
"presence” that is characteristic of such 
a design. 

Detectors using diodes have no gain. 
Indeed, they exhibit a loss. Measure- 
ments with mixers of the type shown in 
Fig. 9A (using two diodes) frequently 
show a low of 5 to 6 dB. The circuits 


using four diodes typically have a 6- or 
7-dB conversion loss. In the high- 
frequency region, and usually through- 
out vhf, the diodes contribute essen- 
tially no noise, making the noise figure 
of such a mixer merely its conversion 
loss. The noise figure of a direct- 
conversion receiver using this as the 
detector will be the mixer conversion 
loss, plus filter losses, plus the noise 
figure of the audio amplifier. It is easy 
to build audio amplifiers with noise 
figures under 3 dB. Hence, receivers 
using direct conversion can be con- 
structed easily to display a respectable 
13-dB noise figure when using diodes as 
the detector. 

Shown in Fig. 10 is a simple direct- 


A 





(C) 


Fig. 9 — Examples of three diode detectors. 


L 


Fig. 10 - A basic direct -conversion receiver using a 5-pole high-pass network at the input port. 


conversion receiver which was as- 
sembled in order to perform some de- 
fector measurements. The input filter is 
iS-pole high-pass type with a cutoff at 
3 MHz. This filter was inserted in order 
lo eliminate a trace of broadcast-band 
rectification which was present. How- 
ever. this was the only selectivity ele- 
ment which was used in the receiver. 
The detector was tire simple two-diode 
type discussed above. It was followed 
by a high-gain audio amplifier, using 
three inexpensive (but “hot”) tran- 
sistors. The diodes were silicon 
switching types (1N914 or equivalent) 
which were matched for forward re- 
sistance with a VOM. A BFO energy of 
•13 dBm (20 milliwatts) was supplied 
irom a homemade general-purpose sig- 
nal generator. 

The first experiment performed was 
to evaluate the sensitivity. Since a min- 
imum of audio filtering was included in 
the system, a careful sensitivity mea- 
surement would not have been very 
enlightening because of the wide band- 
width of the system. However, a signal 
of 0.1 jrV was easily detected at 7 MHz, 
and a l-/iV signal was plainly audible. 
\n audio output of 1-volt rms was 
measured for an input of 6 /uV, in- 
dicating a net receiver gain of 88.4 dB. 

The next measurement was to eval- 
uate receiver blocking. This was done 
with two signal generators and a hybrid 
combiner. The desired signal was from a 
low-level crystal-controlled generator 
which was well shielded. It was set for 
in output of 1 /aV, and the BFO was 

S ' lsted for copy of this signal. The 
er generator was a URM-25D - 
mother well-shielded instrument. This 
was set initially at 50 kHz away from 
the desired signal, and the level was 


increased until blocking occurred. How- 
ever, the measurements were misleading, 
for there is essentially no selectivity 
following the detector except for a 
capacitor which provides low-pass fil- 
tering. The audio amplifier was 
overloading, so die second generator 
was set to 8 MHz. and the experiment 
was repeated. Note that the input was 
broadband in nature. That is, diere was 
no selectivity ahead of die detector. 
Nonetheless, the detector was able to 
provide solid copy of die l-/iV desired 
signal, with no desensitizadon from an 
undesired input signal of 0.1 volt. There 
are many well respected commercial 
receivers which cannot pass this test! 

In spite of die good response to the 
weak and strong signals described, diode 
detectors have deficiencies which make 
diem difficult to use: Diode mixers, in 
general, should be terminated carefully 
if optimum signal-handling ability is to 
be retained. Specifically, the "signal" 
port of the mixer should look back at a 
source impedance of around 50 ohms. 
Further treatment of termination is 
presented during the mixer discussion in 
chapter 6. 

Another characteristic which can 
present a problem, but can be an asset, 
is a tendency toward harmonic mixing. 
Even if the BFO energy supplied to a 
mixer is free of harmonics, the 
nonlinear nature of the diodes will 
create large harmonic currents. The re- 
sult is that input signals at other fre- 
quencies will also cause major outputs. 
Diode balanced mixers are known for 
their high response to odd -order har- 
monics. The receiver of Fig. 1 0 was used 
to evalutc die harmonic mixing traits of 
a simple two-diode product detector. 
The BFO was set at 7 MHz, and die 


signal generator was adjusted to various 
harmonic frequencies, with the audio 
output always adjusted for 1-volt rms. 
The results are presented in Table 1 . 
The dominance of odd-order responses 
is clear from the data. 


Table 1 


N 

Fin 

V in 

Ratio 

1 

7 MHz 

6 jiV 

OdB 

2 

14 

700 

41.3 

3 

21 

20 

10.5 

5 

35 

70 

21.3 

7 

49 

100 

24.4 


The harmonic-mixing phenomenon 
could be used to advantage. For 
example, it might be possible to con- 
struct a receiver which used both the N 
= I and N = 3 responses to cover the 7- 
and 21 -MHz bands. More often, how- 
ever, harmonic mixing is a problem. 
This is especially true if die user lives 
close to commercial TV and fm stations. 
As die receiver is tuned, "birdies” may 
appear across the band. 

The answer to the harmonic-mixing 
problem is, of course, preselection. A 
good low-pass filter ahead of the 
receiver will attenuate harmonic inputs 
to the point that all spurious responses 
are eliminated. This can be more dif- 
ficult to do than might be suspected, for 
it is required that the filters ahead of 
the receiver have the desired attenuation 
not only near die cutoff frequency, but 
in the vhf stop-band. This means that 
vhf layout and shielding methods should 
be used even in a 40-meter filter! 
Filtering of the BFO will do little, for 




Receiver Design Basics 


75 




the mixing harmonics are created in the 
detector. 

A partial solution is to replace the 
silicon switching diodes with hot-carrier 
diodes. These units differ from the usual 
PN semiconductor diodes. They consist 
of a junction between a semiconductor 
(usually N type) and a metal. These 
diodes switch fast and work well 
through the microwave frequencies. 
Furthermore, they lack the charge- 
storage effects which partially cause 
junction diodes to create high -order 
harmonics. 

While harmonic mixing is a major 
problem with diode product detectors, 
it is present to some extent in other 
detectors as well. For example, the 
square-law response of the dual-gate 
MOSFET makes this device prone to 
even-order harmonic mixing. 

A useful attribute of harmonic 
mixing is that it aids the calibration of 
direct-conversion receivers. For 
example, if a 100-kHz oscillator is used 
to calibrate a 7 -MHz direct-conversion 
receiver, it is often possible to hear the 
2nd and 4th harmonics. They can be 
used as markers (for free) at 50- and 
25-kHz intervals. 

The need for preselection filtering is 
significant for the reasons outlined 
above and in the preceding section (e.g.. 
image rejection in superhets). Harmonic 
responses can be suppressed with the 
half-wave low -pass filters described in 
chapter 4. A number of narrow-band, 
multi section band-pass receiver filters 
are presented in the appendix. 

Audio Amplifiers for 
Direct-Conversion Receivers 

Direct-conversion receivers differ in 
a number of ways from the “superhet.” 
Most significant is where the incoming 
signal is detected immediately with no 
intermediate heterodyning processes. 
Another difference is the gain dis- 
tribution. The typical superhet will have 
most of the gain concentrated in the i-f 
section, with only 30 to 60 dB being 
achieved at audio frequencies. On the 
other hand, the direct-conversion re- 
ceiver has nearly all of the gain con- 
centrated in the audio section. Indeed, 
when a diode type of product detector 
is used without an rf amplifier (as 
described in the previous section), the 
only gain in the receiver is in the audio 
stages. 

The high gain requirement of the 
audio section of a direct-conversion 
receiver places more stringent require- 
ments on the amplifier design than 
would be the case with a superhet. Not 
only must the gain be high, there should 
be no instability in the amplifier. While 
oscillations rarely occur in the low -gain 
amplifiers used in superhets, they can 
take place when the amplifier has up to 
100 dB or more of gain. Finally, the 

76 Chapter 5 


noise figure of the audio amplifier is 
significant, especially when low-gain de- 
tectors arc employed (such as those 
using diodes). 

Shown in Fig. 1 1 is a three-stage 
amplifier using 2N3565s. These tran- 
sistors are inexpensive, have high beta 
and low noise figure. Using an amplifier 
design, we will present a fairly detailed 
analysis of this circuit. The transistor 
model is simple. A beta of 100 is 
assumed for each of the transistors, and 
the emitter-base offset is 0.7 volt. A 
10-volt supply is used, and the output 
termination is a set of 2000-ohm head- 
phones. 

The first step is to evaluate the 
biasing of the amplifier. Three direct- 
coupled stages are used. Hence, the 
overall amplifier will be inverting, there- 
by allowing us to use negative dc feed- 
back to bias die circuit. Since all of the 
transistors will be operating in an active 
condition, the voltage on the base of Q1 
will be 0.7 volt. This voltage can orig- 
inate only from the bias resistors from 
die collector of Q3. Noting die values 
used, we see that 0.7 volt occurs at the 
base of 01 only when die dc potential 
at the collector of Q3 is 6 volts. 

Knowing the dc output voltage, we 
can evaluate all of the dc voltages in the 
amplifier. The collector current for Q3 
must be 2 mA [(10 — 6V)2kJ2j, 
leading to V e 3 = 0.2 volt and V b 3 = 0.9 
volt. Continuation of this analysis gives 
us the voltages and collector currents 
for Q1 and Q2. These are shown in 
squares in the figure. 

The next step is to evaluate the 
input resistances for each stage. ForQl 
and Q2, the input resistance of each is 
given by R,„ = 250 -r / e (mA) leading to 
input resistances of 5 kO and 3 kS2 for 
Q1 and Q2, respectively. The input 


resistance of the overall amplifier will be 
the input resistance of Q1 shunted by 
the bias resistors, leading to an overall 
input resistance of rouglily 3 kO. The 
input resistance of Q3 is not given by 
the same formula as was used for the 
first two stages, since emitter degenera- 
tion is used. In this case, R m = PR e is a 
suitable approximation, leading to 
Ri„- i = 10 k 12. 

Having this information, the small 
signal ac gain of the amplifier may be 
calculated. These calculations are pre- 
sented next, assuming a l-/iV input 
signal: K /f , = 1 = V in -f R in , = 2 

X 10“ 0 A, l cl = P /„, = 2 X 10'* A. 
and V eX = l cl R, , = 2 X 10'® X 2.3 X 
10 3 = 4.6 X 10 ' S V. (Note: The col- 
lector load is Rin — 2 paralleled with the 
10-kl2 load resistor.) Next. I b2 = K cl * 
Rj„ 2 = 4.6 X 10" 5 4- 3 X 10 3 = 1.53 X 
10-* A, I c2 =PI b 2 = 1-53 X 10' 6 A. 
V c 2 = I c2 Ri 2 = 7.67 X 10' 3 V. and 
V c3 = G Vi V C 2 = 7.67 X 10' 2 V. Note 
that the emitter degeneration in the last 
stage leads to a voltage gain of 10 in 
that stage. 

The overall voltage gain of the am- 
plifier is 7.67 X 10 4 . Taking 20 Log G v , 
we arrive at 97.7 dB, a value quite close 
to that measured. These methods may 
be used to evaluate any of the simpler 
audio amplifiers which are used in sim- 
ilar applications. 

There are a few subtleties to the 
design of the amplifier of Fig. 1 I . First 
is the 100-ohm emitter-degeneration re- 
sistor in the last stage. This serves a 
number of functions. First, it decreases 
the gain to a level which is compatible 
with the desired overall gain. Addition- 
ally, since the output signals from Q3 
may be large, it adds linearity to this 
stage in order to minimize distortion. 
Finally, it increases the bandwidth of 


r 



Fig. 1 1 - A three-stage, high-gain audio amplifier which uses inexpensive bipolar transistors. 


■ 



Fig. 12 — Circuit and frequency-response characteristics of a passive audio filter. 




the overall amplifier. This is of sig- 
nificance in stabilizing the operation of 
the gain block, because of the dc feed- 
back method of biasing. 

A 0.1 -qF capacitor is shown from 
the base of Q3 to ground. This capacitor 
will have an impedance of about 1 .6 kf2 
at 1 kHz, leading to a low-pass charac- 
teristic for the overall response. Note 
that this impedance is much less than 
the collector load of 5 k£2 on Q2. 

The input impedance of the overall 
amplifier is about 3 kS2. Hence, if the 
input were driven directly from the 
lew -impedance output of a diode type 
of product detector (typically around 
50 ohms) very little of the output 
energy would be transferred. To realize 


driven from a 50-ohm source, and the 
output is terminated in 3000 ohms. 
Note that over 40 dB of attenuation is 
present at 10 kHz. 

Another convenient means for 
achieving high gain at audio frequencies 
is through the use of IC op amps. Most 
of the commercially available op-amp 
ICs have extremely high open-loop gain 
at dc, and are applied easily in audio 
circuits. Considerable care must be used 
if optimum results are to be obtained. 

Shown in Fig. 13 is an audio am- 
plifier using a bipolar transistor and a 
741 op-amp. The advantage of this 
circuit is that it is decoupled easily from 


the supply while still providing high 
gain : In this case about 78 dB (assuming 
the output is terminated in a resistance 
equaling tire input resistance of the 
transistor). The gain of the op amp is 
determined by the feedback resistors, in 
this case the 47-kS2 and l-kf2 units. It 
would be possible to increase the gain 
considerably by shorting the 1 -kS2 re- 
sistor, tli us biasing the op amp to 
operate at its open-loop gain value. 
However, the noise would probably be 
intolerable. If op amps are used in 
high-gain applications, it would be wise 
to use low -noise types. The LM-301A is 
preferred over the 741, and the 
LM-308N is probably one of the best 
low -noise units available. 

While op amps have appeared 
frequently as audio amplifiers in the 
ham literature, they have often been 
misused. The advantage of using an op 
amp over other kinds of circuitry is that 
the performance of the ultimate circuit 
is controllable through the use of feed- 
back. Generally, an op amp should not 
be used in an open-loop manner. Fur- 
thermore, potentiometers should never 
be necessary to bias an op amp in an 
audio application! 

The two amplifiers described in the 
foregoing text are suitable as the major 
gain blocks in many direct-conversion 
receivers. There are other ways to ob- 
tain the needed gain, leaving plenty of 
room for experimentation. The am- 
plifiers described are merely examples. 

If a loud speaker is driven instead of 
2000-ohm headphones, other circuits 
must be used, ones which are capable of 
driving lower impedances. 

Practical Audio Amplifiers 

Integrated circuits have come to the 
fore in recent years, filling a need for 
compact low-power audio amplifiers of 
the transformerless variety. For most 
amateur applications a chip in the 025- 
to 2-watt class is suitable. The majority 


the full gain of the amplifier, an im- 
pedance transformation is required at 
the input. This could be a simple audio 
transformer with a turns ratio of, say, 
1:5, Transformers at the input to a 
high-gain block of this kind are often 
difficult to use owing to their tendency 
to pick up 60-Hz energy. Shown in Fig. 
12 is an alternative solution. Here, an 
88-mH toroid is used as the inductor in 
an L network. A pot core could be used 
if a toroid was not available. This 
network has a peak response at 940 Hz, 
where the impedance transformation is 
well over 10. As an additional bonus, 
the L network serves as a low -pass filter, 
offering protection to the audio am- 
plifier from out-of-passband signals. The 
figure also shows a computer-calculated 
response of this filter when the input is 



Fig. 13 - An audio amplifier capable of 78 dB of gain. It combines a bipolar transistor and 
an op amp. 


' 


Receiver Design Basics 


77 






Fig. 14 — Examples of audio amplifiers. 


of these ICs are designed to operate into 
a nominal load impedance of 8 ohms at 
the rated harmonic distortion character- 
istic set by the manufacturer. However, 
headphones can be substituted for a 
speaker in most instances, regardless of 
the headset impedance (4 to 2000 
ohms), and satisfactory operation will 
result without damage to the IC re- 
sulting from a mismatched condition. 

One problem exists when certain 
audio ICs are used: Biasing is done 
internally, thereby preventing the 
builder from improving the cross-over 


distortion characteristics. Distortion of 
that kind is not especially troublesome 
at high audio-output levels, but during 
weak-signal reception, and at mod- 
erately low audio-output amounts, the 
distortion will affect the quality of the 
received signal. A cw note, for example, 
will exhibit a fuzzy sound which can 
impair readability. 

The use of discrete devices in an 
audio-output stage (at power levels 
above, say. 100 mW) permits the de- 
signer to tailor the circuit for minimum 
cross-over distortion. It would be waste- 


ful in a serious design effort to have a 
high-performance rf/i-f receiver section, 
then degrade the signal quality by em- 
ploying a substandard audio channel. 
The linearity of all the stages in an 
audio system should be as good as the 
art will permit. At least, the designer 
should strive to meet that criterion. 

Attention must be paid to the audio 
voltage levels entering each af stage at 
maximum signal amounts. That is, the 
amplification capability of each stage 
should be set so that a successive stage 
will not be driven into a nonlinear state. 
Gain distribution is as significant as it is 
in the early stages of a well-designed 
receiver. Also, the frequency response 
of the stages should be shaped for the 
desired audio passband characteristics. 
This subject is treated elsewhere in the 
book. The high-frequency response of 
the audio system should roll off at the 
highest desired frequency - typically 
1000 llz for cw work, and 2500 Hz for 
ssb reception. The net effect is one of 
minimizing high-frequency noise and 
heterodynes. This aids in reducing the 
QRM problem and enhances the overall 
signal-to-noise characteristics of the re- 
ceiver. Some cw operators prefer an 
even lower roll-off point for the audio 
system — 600 or 700 Hz. Similarly, one 
may desire to cause a low-frequency 
roll-off in tire 100- to 300-Hz region. 
The exact frequency is a matter of 
subjectivity, depending on the op- 
erator's choice of receiver fidelity. A 
good low-frequency roll-off will im- 
prove reception by eliminating much of 
the low-frequency rumble caused by 
QRN and sideband energy from ssb 
stations operating near the chosen fre- 
quency. Furthermore, 60-Hz hum prob- 
lems are minimized if shaping of that 
kind is used. Low-impedance hi-fi head- 
phones are not recommended for use 
with receivers which do not have audio 
systems that have been shaped for com- 
munications bandwidths. The wide fre- 
quency response of such headsets will 
degrade the readability of weak signals 
by allowing noise and high-pitched het- 
erodynes to pass, to say nothing about 
60- and 120-11/ hum that may be 
present. 

In the interest of reducing the har- 
monic distortion level of an audio- 
output amplifier, it is useful to have 
more audio power capability than is 
required. When the maximum rated 
power of an audio IC or discrete-device 
amplifier is depended upon for adequate 
sound level, the system is operating in 
its maximum harmonic-distortion 
region. Hi-fi designers rely on the 
general concept of having more audio 
than is needed, thereby permitting the 
amplifier to operate over a portion of its 
curve where minimum distortion will 
occur. 

A 0.5-W IC audio amplifier is shown 


78 Chapter 5 









Fig, 15 - Example of a two-pole passive audio filter which contains an 88-mH toroidal-wound 
inductor in each resonator. 


in Fig. 14A. A Motorola plastic 8-pin 
dual inline device is used. The chip 
contains a preamplifier and audio- 
output section for driving an 8-ohm 
load. The preamplifier voltage gain is 
nominally 100, and the audio power 
amplifier has a gain of 10. Tire com- 
bination provides a voltage gain of 
1000. With 3 mV of input signal, 0.5 W 
of audio output will occur. 

No-signal resting current is approx- 
imately 4 m A at 9 V. The IC works 
nicely with headphones In the 8- to 
2000-ohm impedance class and is quite 
suitable for use in small portable re- 
ceivers. The 33-/iH rf choke seen at the 
output port is used to suppress hf 
parasitic oscillations which can occur. 
Such unwanted energy can radiate from 
the circuit board and speaker leads, 
causing interference to the front end 
and i-f sections of a receiver. For op- 
eration from a 12- or 13-V power 
supply, it is a simple matter to drop the 
1C operating voltage to 8 or 9 volts by 
means of a three-terminal regulator. If 
the 1C is operated from a 9-V battery, a 
300-j/F capacitor should be placed in 
parallel with the battery to prevent 
distortion caused by increased battery 
resistance as the battery becomes de- 
pleted. Under normal operating con- 
ditions die harmonic distortion is rated 
at 0.5 percent at 250 mW of output to 
an 8 -ohm load. 

A 1-Watt Amplifier 

A Motorola MC1454G can be used 
when a power output of I watt is 
desired. The IC has ten leads and is 
contained in a 602B style case (similar 
to a TO-5 case). Total harmonic dis- 
tortion is rated at approximately 0.8 
percent at I kHz while using a 16-ohm 
load. A practical circuit is given in Fig. 
1413 . Zero-signal current is ap- 
proximately 1 1 mA. 

The diagram shows the IC config- 
ured for an A y (voltage gain) of 18, but 
by making minor changes in the pin 
connections one can set the gain at 10 
or 36. depending on die operator's 
requirements. Details are given in die 
Motorola data sheet. 

Networks consisting of a 10-ohm 
resistor and a 0.1 -/a F capacitor are con- 
nected to ground from pins 9 and 10. 
They help to prevent unwanted rf oscil- 


lafions. The R-C networks and all other 
circuit connections to the chip should 
be kept as short as possible to ensure 
stability. A .05-/aF capacitor is em- 
ployed between pin 1 and ground to 
decrease the amplifier bandwidth 
another aid to stability. This IC can be 
used safely with headphones which ex- 
hibit impedances from 4 to 2000 ohms. 
Similarly, a 4- or 8-ohm speaker can be 
used in place of a 16-ohm one, but the 
lower the voice-coil impedance below 
16 ohms, the greater the percentage of 
harmonic distortion. 

A 3.5-Watt Amplifier 

In applications where maximum cur- 
rent drain is not a matter of prime 
importance, the circuit of Fig. 14C is 
worthy of consideradon. A com- 
plementary-symmetry Class B audio 
pair, Q1 and Q2, is driven by U3, a 
noninverting voltage amplifier which 
serves as a phase splitter. 

This circuit is designed to deliver 
approximately 3.5 watts to a 4-ohm 
load. Supply voltage can range from 12 
to 14. THD (total harmonic distortion) 
will be roughly 0.25 percent at 3.5 
watts output. Most of the voltage gain is 
effected at U3, with Q1 and Q2 rep- 


resenting a voltage gain of 3. 

An audio preamplifier is necessary 
ahead of U3 if the system is to be used 
directly after a product detector. A 
single-stage Class A amplifier, such as a 
2N2222A. will suffice. R1 functions as 
a protective circuit for the input of U3 
during discharge periods of Cl. CR1 
serves as an antisaturation clamp to 
prevent latchup of U3. This circuit is 
patterned after one described by Jung 
(/C Op Amp Cookbook). Idling current 
is practically zero because Ql and Q2 
are biased off during no-signal periods. 
Additional audio amplifiers for driving a 
speaker are presented in the ARRI. 
Electronics Data Book and in the Hand- 
book. 

Audio Filters 

When overall selectivity in a receiver 
is lacking, especially for cw use, a 
significant improvement can be realized 
with the addition of an audio filter. 
There are two common situations. One 
is when a superhet receiver is designed 
primarily for ssb and has an i-f band 
width of approximately 2 kHz. If this 
receiver is used for cw, an audio band- 
pass filter can do wonders in reducing 
the effects of QRM. The other case is 
when the receiver follows the direct- 
conversion concept, where all adjacent- 
channel selectivity must, by necessity, 
originate at audio frequencies. 

Audio filters may be synthesized 
through two methods. The first is where 
inductors and capacitors are used to 
form resonant circuits. These resonators 
are coupled in order to obtain multipole 
responses. The other technique (more 
popular) is the use of R-C active-filter 
sections. Here, capacitors and resistors 
are used in conjunction with feedback 



Fig. 1 6 — Examples A and B show methods for terminating an 1C filter. 


Receiver Design Basics 79 






Fig. 17 - A simple low-pass filter section Fig. 18 - Curves tor output voltage versus input frequency, illustrating the effects of Q 

using an active device. 


amplifiers in order to synthesize the 
same effect that could be obtained with 
a passive combination of inductors and 
capacitors. The advantages of the latter 
are many. First, inductors for the audio 
frequencies are bulky, heavy and ex- 
pensive. Their losses are often high. 
Conversely, resistors and capacitors are 
lightweight and compact, and are inex- 
pensive. If desired, gain can be obtained 
from an active filter. 

Shown in Fig. 15 is a simple two- 
pole band-pass filter which is designed 
around an 88-mH toroidal inductor of 
the kind used by RTTY enthusiasts. 
This filter was designed (using predis- 
torted Butterworth tables) for a center 
frequency of 800 Hz and a 3-d B band- 
width of 150 Hz. The measured un- 
loaded Q of the inductors was approx- 
imately 25 at 1 kHz. 

The operation of any l.C selective 
filter is critically dependent upon the 
resistive terminations at each end of the 
filter. The unit described in Fig. 14 
must have a termination of 4.7-kS2 on 
each side if the proper passband is to 
result. Shown in Fig. 16 are two suitable 
methods for terminating the l.C filter. 
Bodi of these systems can provide con- 
siderable gain. In the case where op 
amps are used, the designer should 
remember that the use of feedback 
causes both the output impedance and 
the impedance looking into the in- 
verting port to be essentially zero. 

The more exciting technique for 
audio filter design is the R-C active 
approach. Virtually all of tire response 
types of interest can be handled. This 
includes the low -pass, high-pass, and 
bandpass responses as well as assorted 
band-reject and all-pass functions. An 
example of an all-pass response would 
be seen in the phase -shifting networks 
of the kind used in phasing-type ssb 
transmitters or receivers. Only simple 


low-pass and band-pass responses will be 
considered in this section. 

Shown in Fig. 1 7 is a simple low-pass 
filter section. This circuit should be 
driven from a low-impedance source 
one with an output resistance much less 
than the R used in the filter. Aide this 
circuit will have a voltage gain of unity. 
However, at well above the cutoff fre- 
quency there will be significant atten- 
uation. The response near the center 
frequency will depend upon die design 
Q of die network, which is determined 
by die ratio of the two capacitors used. 
The output voltage will be Q times the 
input voltage at the center frequency, 
fo- Fig. 18 presents curves of output 
voltage versus input frequency for cases 
where Q is 1/2. 1.3 and 5. 

The amplifier used for filters of this 
kind is quite simple. The voltage gain 
should be unity and the amplifier 
should be noninverting. A simple emit- 
ter follower using a high-beta transistor 
such as die 2N3565 is often suitable. 
Shown in Fig. 19 are two other circuits 
which may be used. One is a 741 or 
similar op amp, wired in the follower 
configuration. The other uses a pair of 
transistors in a feedback arrangement. 
Both amplifiers should be biased so the 
dc voltage is approximately half the 
supply voltage. 

Useful filters are built using the 
circuits just discussed by cascading 
many sections. The fact that this circuit 
has unity gain at dc makes biasing easy. 
An example is shown in Fig. 20. The 
first unity gain amplifier is used as a 
follower to bias die following stages 
properly. The 10-qF input capacitor is 
large enough to allow response down to 
low frequencies. A 0.1 -qF unit would be 
desirable since this would cause the 
input section to act as a single-secdon 
high-pass filter. This would ensure con- 
siderable attenuation at 60 and 120 Hz. 


Earphones can be driven directly from 
the outputs through an electrolydc cap- 
acitor. 

In principle, any number of filter 
sections may be cascaded to obtain the 
response desired. For most amateur 
applications identical filter sections are 
used, resulting in a Bessel type of 
transfer response, while simplifying the 
design procedure. It is not necessary 
that the sections be identical. If the 
cutoff frequencies and individual 
section Qs are chosen properly. Butter- 
worth and Chebyshev response filters 
may be synthesized. 

Shown in Fig. 21 is a single-section 
band-pass filter. This circuit differs from 



80 Chapter 5 







the low-pass one because there is no 
response at dc. and the attenuation at 
high frequencies is not as pronounced as 
with the low -pass filter. The filter offers 
sonic simplification because the capac- 
itors are equal in value. Furthermore, 
this circuit is capable of yielding con- 
siderable voltage gain at die center 
frequency. 

Shown in Fig. 22 are normalized 
voltage responses for diis circuit, as a 
function of frequency, for design 0s of 
1, 3 and 5. The voltage gain at the 
center frequency can be as high as 2 0 2 . 

While high voltage gain is sometimes 
an advantage, it can cause a problem if 
the filter is used with an existing re- 
ceiver. in such cases.it is more desirable 
to operate a filter with a gain close to 
unity, or just slightly above. The band- 
pass circuit of Fig. 21 is modified easily 
by including an attenuator section at 
die input, which causes die overall 
voltage gain to be ll 0 . This is any 
desired value less dian or equal to the 
maximum available value of 2 Q 2 . 

Since the filter section of Fig. 23 has 
no output response for a dc input signal, 
it requires a different approach to 
biasing if a single power supply is used. 
A circuit using several bandpass sections 
with a single power supply is shown in 
Fig. 24A. Multisection filters of this 
kind may be built with op amps, such as 


the 741, 747, ’5558 duals or the 
LM-30I A. For critical low-noise applica- 
tions the LM-308N would be ideal, but 
it is more expensive. 

Other circuits may be employed to 
obtain a bandpass response. However, 
the results would be essentially die 
same. The simple band-pass section dis- 
cussed has the advantage that it is not as 
sensitive to component variations as 
some other circuits. This general ap- 
proach is used commercially for some 
ready-built filters offered to the radio 
amateur. 

Both of the R-C aedve filters pre- 
sented allow latitude to the designer in 
the choice of components. In each case 
the capacitors may be picked on an 
arbitrary basis. The design frequency 
and the 0 are then chosen. For the 
low-pass filter the 0 will place a con- 
straint upon the ratio of the capacitors, 
while the center-frequency gain must be 
chosen for the band-pass case. After 
these parameters are pinned down, the 
resistor values can be calculated. For 
low-0 situations (0s less than 6 or 8), 
the nearest 10-percent resistors can be 
used. It is advisable to select the larger 
capacitance values, for this leads to 
lower resistance values, and keeps the 
impedances low enough to maintain a 
low -noise output. Miniaturization would 
lead one in the opposite direction. For 


the low-pass filter, a value of 0.1 /jF for 
Cl is a good starting point, with C2 
being picked to yield the desired section 
0. A value of .022 n? is suitable for the 
band-pass circuit. 

Care must be used when applying 
these ideas to the design of a direct- 
conversion receiver. Ideally, for best 
dynamic range, the place for selectivity 
in any receiver is at as low a signal level 
as possible. However, noise considera- 
tions may not allow this route to be 
followed. For example, the active band- 
pass filter discussed has a resistive at- 
tenuator at its input if it is designed for 
anything less than maximum possible 
gain. This attenuator, along with the 
noise in the op amp used for the first 
filter section, would severely com- 
promise the noise figure and sensitivity 
of a receiver which used a diode type of 
product detector — if the filter were to 
follow the detector. On the other hand, 
if all of the selectivity of a direct- 
conversion receiver was concentrated at 
the output of the audio amplifier, one 
would have an acceptable noise figure, 
but the audio amplifier would severely 
overload from adjacent-channel signals. 
The best approach would be a com- 
bination of the two methods. That is, 
some passive low-pass filtering should be 
used between the product detector and 
the first audio amplifier in order to 
protect the audio amplifier, with the 
major close-in selectivity achieved after 
some amplification. It is worthwhile to 
include selected capacitors within the 
audio amplifier to attenuate the higher 
audio frequencies. 

A question often posed is whether to 
use a low-pass or a band-pass filter. This 
query is difficult to answer, for it will 
depend to a large extent upon the 
personal preferences of the user. Cer- 
tainly, the sharp band-pass filter built 
with four or five sections, each having a 




Fig. 77 - Curves for output voltage versus input frequency of the single-section band-pass 
filter. 


Fig. 21 - Representation of a single-section 
active band-pass filter. 


Receiver Design Basics 81 






Fig. 23 - Band-pass filler with suitable de- 
sign equations. 


Q of 5, will be impressive. However, 
such a filter can cause mental fatigue if 
it is used for long periods, such as 
during contest operation. 

The writers feel that a low-pass filter 
with a cutoff frequency of roughly I 
kHz, but with several sections to ensure 
attenuation at high frequencies, is su- 
perior for use with most direct- 
conversion receivers. Such a tiller is 
shown in Fig. 24B. The constants for a 
ssb unit are also included. Each section 
is designed with a Q of unity. However, 
two low -value coupling capacitors are 
used at the input and between the last 
filter section and the low-gain output 
amplifier in order to attenuate low 
frequencies and hum. The latter can be 
troublesome with direct-conversion re- 
ceivers. This filter has been used with a 
number of the direct-conversion re- 
ceivers and transceivers described in this 
book. Pleasing results were had. 

An ideal solution would be to in- 
clude both filter types in a receiver. The 
low-pass filter of Fig. 24 could be 
followed by a bandpass unit with a 
center frequency of 800 Hz and a 
narrowband-width. This filter, probably 


containing only two or three sections, 
could be used when necessary. 

Superhet Basics - I-F System 
and Filter Design 

In the first section of this chapter, 
the basic ideas governing the design of a 
superhet receiver were presented and 
were contrasted to direct-conversion de- 
signs. Now. some design information is 
presented concerning the general 
methods to be used in designing the i-f 
section of a superhet. This includes a 
discussion of crystal filters and other 
methods for obtaining selectivity. In the 
next section, the details of some dif- 
ferent approaches for building and 
analyzing suitable amplifiers will be 
presented. 

Envision a superhet receiver which 
was typical of those used in the late 
1940s and early 1950s. This unit was a 
single-conversion variety — the incoming 
signal was applied to a mixer, then 
converted to an i-f where the main 
selectivity and gain of the receiver was 
obtained. Then, the signal was detected, 
yielding audio which was further am- 
plified and applied to headphones or a 







speaker. The usual i-f was 455 kHz. 

Such a receiver, set for reception at 14.0 
MHz, is seen in Fig. 25. 

Note that the local oscillator in this 
receiver is operating at 14.455 MHz in 
order to produce a 455-kHz i-f from an 
arriving 14-MHz signal at the antenna 
terminals of the receiver. However, the 
i-f image in such a receiver is the other 
incoming signal at the mixer input 
which would also provide a 455-kHz 
output: in this case. 14.910 MHz. To 
keep the receiver from being dominated 
by these image responses, extensive 
front-end filtering is required. The 
filtering ahead of the mixer should be so 
selective that the 14-MHz signal is 
passed with minimal attenuation while 
offering considerable attenuation to sig- 
nals at 14.91 MHz. Such filters can be 
designed easily, but they are not easily 
realized in a receiver which must tune 
over large frequency ranges. Many re- 
ceivers , of the early 1950s had two 
tuned circuits which were separated by 
an rf amplifier, yielding 40 to 50 dB of 
image rejection during 20-meter opera- 
tion. On die lower amateur bands, die 
image rejection was better, although up 
on 10 meters, the image rejection was as 
little as 10 or 20 dB. 

This image -rejection problem led to 
the popularity of dual -con version re- 
ceivers. The early units were similar to 
that shown in Fig. 26 where the 
incoming signal was converted first to 
an i-f of roughly 2 MHz, then was 
converted to a lower second i-f. The 
latter was often at 455 kHz. although 
many units used lower frequencies 
where selective transformers were more 
easily constructed. Triple-conversion re- 
ceivers were used also. A third i-f of 50 
kHz was popular. 

A second form of dual-conversion 
receiver was built by Collins Radio (Fig. 

27). The first local oscillator was crystal 

controlled. The first i-f, typically Fig. 26 — Representation ot a dual-conversion superheterodyne receiver, 
around 2.5 MHz in amateur receivers for 
the 3- to 30-MHz region, was a broadly 
tuned affair, often widi a bandwidth 
from 200 to 500 kHz. This broad first 
i-f was converted to a selective second 
i-f. The advantage of dtis scheme was 
that the stability of the receiver was 
excellent because of die crystal- 
controlled first-conversion oscillator. 

Good frequency accuracy resulted from 
the high precision which could be used 
in designing the second oscillator. This 
was possible since only one tunable 
oscillator was required. 

The image -rejection ratio of dual- 
conversion receivers of this vintage was 
often 60 to 80 dB. aldiough this was 
rarely reflected in the conservative spec- 
ifications offered by the manufacturers. 

Moreover, this image rejection was 
usually as good on the 10-meter band as 
it was on 80 or 40 meters. 

In spite of improved image rejection Fig. 27 - Dual-conversion receiver format used bv Collins Radio Co. 

Receiver Design Basics 83 





Fig. 25 - Typical receiver format for the late '40s and early '50s. 














and stability, the dual -conversion re- 
ceivers outlined often have problems. 
These are related to the incoming signal 
being subjected to several stages of 
amplification prior to “seeing" the 
highly selective filters which would ap- 
pear in the final i-f system. When an 
incoming signal is subjected to several 
stages of gain, it glows to fairly high 
levels. This means that effects from 
nonlinearity can become significant. 
These include cross modulation, inter- 
modulation distortion, and blocking. 
These effects ‘will be discussed in the 
next chapter. 

A partial solution to the nonlinearity 
problem lies in the use of a single- 
conversion receiver design, as depicted 
in Fig. 28. This receiver, which rep- 
resents most modern units used by 


today’s amateur, differs from the classic 
single-conversion receiver in that a 
highly selective filter, usually based 
upon a multiplicity of high-(2 quartz 
crystals, is used at the input to the i-f 
amplifier. This filter is usually the most 
selective circuit in the receiver, and 
serves not only the purpose of defining 
the overall adjacent -signal selectivity of 
the receiver, but of protecting the 
following i-f circuit from strong out-of- 
passband signals. In such a design, only 
those stages preceding the i-f filter are 
significant in producing the nonlinear 
effects which lead to cross modulation, 
IMD, and blocking by out-of-passband 
signals. The design of the front end of a 
superhet will be considered later. 

The image rejection of a single- 
conversion receiver of this sort may still 



Fig. 28 — Representation of modern-day single-conversion superheterodyne receiver. 



Fig. 29 - Illustration of how a mechanical filter operates. 


be excellent. For example, the receiver 
shown in Fig. 28 is for reception of the 
28-MHz band. The i-f is 9 MHz and the 
local oscillator is at 19 MHz. In this 
case, the image frequency is 10 MHz. 
Building a front-end preselector filter 
which will offer significant attenuation 
to 10 MHz (when tuned to 28 MHz) is 
routine. Image-rejection ratios to 60 to 
100 dB are obtained easily. With a 
9-Mllz i-f system the ultimate image 
rejection is often limited not by the 
design of the preselector filter itself, but 
by shielding and isolation practices. 

This brings us to the meat of this 
section: the design of high-frequency 
crystal filters. The commercial filters 
which are popular among amateurs are 
manufactured in West Germany by 
KVG and marketed in the USA by 
Spectrum International. The reader 
should consult the advertisements in 
QST and Ham Radio for information on 
these filters. KVG filters are offered 
with center frequencies of 9 or 10.7 
MHz. Filters with a center frequency of 
3.395 MHz are available from Heath Co. 
Various crystal filters are offered on the 
surplus market, many with low prices 
and superb specifications. Some surplus 
filters have deficiencies which may de- 
grade their usefulness. Beware! 


Electromechanical Filters 

A component which is useful for 
maintaining the required i-f selectivity 
of a receiver is the mechanical filter. 
Collins Radio Company introduced the 
first production models of this filter in 
1952, and the Japanese followed with a 
similar unit in the mid 1960s (Kokusai). 

Perhaps the most significant feature 
of a mechanical filter is the high Q of 
the resonant metallic disks it contains. 
A Q figure of 10,000 is the nominal 
value obtained with this kind of resona- 
tor. If L and C constants were employed 
-to acquire a bandwidth equivalent to 
that possible with a mechanical filter, 
the i-f would have to be below 50 kHz. 

Mechanical filters have excellent 
frequency-stability characteristics. This 
makes it possible to fabricate them for 
fractional bandwidlhs of a few hundred 



Fig. 30 - Analogous representation of a mechanical filter. 

84 Chapter 5 







Fig. 31 — Examples of series and parallel resonating when using mechanical filters. 


Hz. Bandwidths down to 0.1 percent 
can be obtained with these filters. This 
means that a filter having a center 
frequency of 455 kHz could have a 
bandwidth as small as 45.5 Hz. By 
inserting a wire through the centers of 
several resonator disks, thereby coupling 
them, the fractional bandwidth can be 
made as great as 10 percent of the 
center frequency. The upper limit is 
governed primarily by occurrence of 
unwanted spurious filter responses 
adjacent to the desired passband. 

Mechanical filters can be built for 
center frequencies from 60 to 600 kHz. 
The main limiting factor is disk size. At 
the low end of the range the disks 
become prohibitively large, and at the 
high limit of the range the disks become 
too small to be practical. 

An illustration of how a mechanical 
filter works is given in Fig. 29. As the 
incoming i-f signal passes through the 
input transducer il is converted to 
mechanical energy. This energy is passed 



Fig. 32 — Electrical equivalent of a quartz 
crystal. 


through the disk resonators to filter out 
the undesired frequencies, then through 
the output transducer where the 
mechanical energy is converted back to 
the original electrical form. 

The transducers serve a second func- 
tion: They reflect the source and load 
impedances into the mechanical portion 
of the circuit, thereby providing a 
termination for the filter. An analogous 
representation of a mechanical filter is 
given in Fig. 30. 

Mechanical filters require external 
resonating capacitors which are used 
across the transducers. If the filters are 
not resonated, there will be an increase 
in insertion loss, plus a degradation of 
the passband characteristics. Concerning 
the latter, there will be various 
unwanted dips in the nose response 
(ripple), which can lead to undesirable 
effects. The exact amount of shunt 
capacitance will depend on the filter 
model used. The manufacturer's data 
sheet specifies the proper capacitor 
values. 

Most bipolar transistor i-f amplifiers 
have an input impedance of 1000 ohms 
or less. There are situations where the 
output impedance of the stage pre- 
ceding the filter is similarly low. In 
circuits of this variety it is best to use 
series resonating capacitors in prefer- 
ence to parallel ones. Examples of both 
methods are shown in Fig. 31. Stray 
circuit capacitance, including the input 
and output capacitances of the stages 
before and after the filter, should be 


subtracted from the value specified by 
the manufacturer. 

Collins mechanical filters are avail- 
able with center frequencies from 64 to 
500 kHz and in a variety of band- 
widths. Insertion loss ranges from 2 dB 
to as much as 12 dB, depending on the 
style of filter used. Of greatest interest 
to amateurs are the 455-kHz mechanical 
filters specified as F455. They are avail- 
able in bandwidths of 375 Hz, 1.2 kHz, 
1.9 kHz, 2.5 kHz, 2.9 kHz, 3.8 kHz and 
5.8 kHz. Maximum insertion loss is 10 
dB, and the characteristic impedance is 
2000 ohms. Different values of reso- 
nating capacitance are required for the 
various models, spreading from 350 to 
1100 pF. Although some mechanical 
filters are terminated internally, this 
series requires external source and load 
terminations of 2000 ohms. The F455 
filters are the least expensive of the 
Collins line. 

Crystal Filters 

Although a complete theoretical 
understanding of crystal filters is com- 
plicated, it is possible for the advanced 
amateur to build his own filters. This 
possibility should not be dismissed as a 
viable approach. We will not describe 
the design procedure from a formal 
point of view: Some basic concepts will 
be presented which should allow some 
filters to be built empirically. 

Shown in Fig. 32 is the equivalent 
circuit for a crystal. It is used as the 
basis for filter synthesis. This circuit 
shows the normal series-resonant circuit 
consisting of tire motional inductance 
and motional capacitance which are in- 
herent in the piezoelectric crystal. The 
parallel capacitance. C p , is predomi- 
nantly a result of the metallic plating 
which is used to provide electrical con- 
nection to the quartz plate. Also shown 
is a series resistance, R s , which repre- 
sents the losses in a crystal. 




FREQUENCY 


Fig. 33 — Test setup for evaluating a quartz 
crystal. 


Receiver Design Basics 85 





Fig. 34 — Simple form of crystal filter with 
phasing trimmer. 


A test circuit to evaluate a crystal is 
shown in Fig. 33. Also shown is the 
response which might be seen if the 
signal generator was swept slowly 
through tire frequency range of interest. 
The highest response is measured at the 
series-resonant frequency, where the 
motional capacitance and inductance 
resonate with each other. The amplitude 
of this response is slightly below the 
dotted line which represents the signal 
seen if the crystal is short-circuited. The 
difference in dB between the series 
response and the response without the 
crystal may be used to calculate die 
value of R s , the series loss resistance. 

The loaded 3-dB bandwidth is also 
shown. This value may be used to 
calculate a loaded Q for die crystal. If 
this is used in combination with die 
insertion loss associated with R s , the 
unloaded Q of die crystal may be 
calculated. Alternatively, the unloaded 
Q of die crystal may be measured 
directly by placing low-value resistors 
(typically just a few ohms) from each 
side of the crystal to ground. Extreme 
signal-generator stability is required for 
this measurement. 

Also shown in Fig. 33 is a parallel- 
resonant frequency, f p . This resonance 
arises from the series combination of 
the motional inductance and capaci- 
tance, which appears to be an inductor 
at frequencies above the series-resonant 
frequency. This inductance, when 
combined with the parallel capacitance, 
C p , forms a "trap” circuit, causing a 
null in the test output at f p . The 
difference between the series- and 
parallel -resonant frequencies is called 
the pole-zero spacing of the crystal. 

The parallel capacitance of the 
crystal. C p , may be measured directly 
while using a bridge operating at fre- 
quencies far removed from the resonant 
frequencies of the crystals. Audio fre- 
quencies are used for this measurement. 

The values which one obtains from 
these measurements are much different 
dian those encountered with classic LC 
tuned circuits. For example, an 80- 
meter crystal was studied while using 
homemade test equipment, leading to a 
motional inductance of 69 mH, a 
motional capacitance of .029 pF, a 
parallel capacitance of about 8 pF, and 
a series resistance of 21 ohms. The 


unloaded Q was 76.000 and the pole- 
zero spacing was approximately 3 kHz. 

The simplest form of crystal filter 
which can be built by the amateur uses 
one crystal, and is shown schematically 
in Fig. 34. A trifilar transformer is used 
(wound on a ferrite toroid core) in 
order to provide push-pull drive. One of 
die outputs drives the crystal directly. 
The other (out-of-phase) is applied to a 
variable capacitor. This variable is 
adjusted for about the same capacitance 
as the crystal parallel capacitance, and 
has the effect of canceling the parallel 
resonance of the crystal, leaving a 
series-resonant circuit. The value of the 
terminating resistance, R,, will deter- 
mine the loaded bandwidth (BWL) of 
the circuit. The greater the resistance, 
the wider the filter will be. This circuit 
is essentially the same as that which was 
used in tire simple crystal filters in 
receivers built before 1960. 



Fig. 35 - Circuit for a half-lattice crystal filter. 


Shown in Fig. 35 is another common 
circuit, the half-lattice filter. The paral- 
lel capacitances of the two crystals tend 
to cancel each other, leaving the 
response of the filter dominated by the 
series resonances of the crystals. The 
transformer consists, usually, of a bifilar 
output winding on a tuned circuit which 
is in the output of a mixer. The crystals 
are on different frequencies. The overall 
bandwidth of the resulting filter is 
approximately 1 to 1.5 times the fre- 
quency separation of the crystals. The 
spacing in frequencies should not ex- 
ceed the pole-zero spacing of the 
crystals, and the crystals should be 
identical except for the slightly 
different frequencies. This kind of filter 
is used in a simple superhet to be 
described later. In building a filter of 
this kind, it will be necessary to experi- 
ment with the terminating resistance. 
Generally, with a high-value terminating 
resistor, there will be passband ripple. 
As tire resistance is decreased, the ripple 
will disappear, leaving a fairly flat 
response over a bandwidth determined 
by the separation in crystal frequency. 

Shown in Fig. 36 is a modified 
version of the filter just described. Four 
crystals are used. This filter is called a 


cascade half lattice. The transformer 
balances the drive to the crystals, al- 
though the input and output are single 
ended. The balancing transformer may 
be built with a few bifilar turns on a 
ferrite toroid. Alternatively, a bifilar 
winding can be used on a powdered-iron 
core. The circuit is resonated with a 
variable capacitor. Y1 and Y4 should 
have the same frequency within a toler- 
ance of 10 or 20 percent of the band- 
width of the filter. Similarly, Y2 and Y3 
should be matched, although these fre- 
quencies will be different from Yl and 
Y4. The bandwidth will be a little 
greater than the frequency difference. 
As was the case with the simple half- 
lattice filter, the terminating resistances 
are critical. They must be adjusted in 
order to minimize the passband ripple. 
This type of filter, and variations of it 
using additional crystals, is the form 
used for many filters currently em- 
ployed for ssb and fm equipment. 

Another form of filter is shown in 
Fig. 37. In this example a four-pole 
filter is presented. In principle this filter 
may use from two up to dozens of 
crystals. This filter is called the “lower- 
sideband ladder” configuration, since 
when it is built for wide bandwidths, it 
has an asymmetrical response which 
tends to pass the lower sideband. Filters 
of this kind are attractive to the ama- 
teur experimenter, for a filter is gener- 
ally built with all of the crystals cut for 
the same frequency. The empirical 
approach is to choose the values of the 
coupling capacitors and terminating 
resistances in order to arrive at the 
desired bandwidth. This can be done by 
the advanced amateur who is willing to 
build some swept oscillators in order to 
perform the alignment. 

Generally, filters using the lower- 
sideband ladder configuration are 
limited to bandwidths which are much 
narrower (50 percent or less) than the 
pole-zero spacing of the crystals. The 
ultimate passband attenuation of such a 
filter will be limited by the ratio of the 
parallel capacitance of the crystals to 
the coupling capacitors. This makes the 



Fig. 36 - Example of a cascaded half-lattice 
crystal filter. 


86 Chapter 5 






Fig. 37 - Details of a 4-pole lower-sideband 
ladder filter. 


configuration more applicable for cw 
bandwidths. As a starting point the 
amateur should consider coupling capac- 
itors up to a few hundred pF and 
terminating resistances of 50 to 500 
ohms. Practical examples of this filter 
are not given here, since the filter 
components are highly dependent upon 
die exact characteristics of the crystals 
used. 

These comments should be kept in 
mind by die home designer. Many of 
these statements apply also to LC fil- 
ters. 

1) The terminating resistances of a 
crystal filter will critically affect the 
response shape and bandwidth. 

2) The bandwidth of a multisection 
filter is determined predominantly by 
the loaded Q of the resonators used and 
is not a strong function of the number 
of resonators used. 

3) The shape factor of the filter 
(bandwidth at 60 dB down, divided by 
the bandwidth at 6 dB) is a function of 
die number of resonators used and 
tends to be invariant with filter band- 
width. 

4) Extreme care should be used in 
mounting a crystal filter in order to 
preserve the ultimate attenuation which 
the filter is capable of exhibiting. Great 
care should be taken to ensure that the 
input of the filter is well isolated from 
the output. The filter, if built in a 


metallic can, should be mounted 
directly against a metallic ground plane. 

Intermediate-Frequency Amplifiers 

The intermediate-frequency (i-f) 
amplifier is a critical section of a super- 
heterodyne receiver. Not only must this 
system provide a large part of the 
overall gain, but it is die place where 
most, if not all, of the gain control of 
the receiver occurs. Both of these func- 
tions must be kept in mind when a 
design is formulated. The noise figure of 
the i-f amplifier is also of some concern, 
although it is certainly not as critical as 
in the front-end part of a receiver. 

Consider a modern superhet as 
shown in Fig. 38. The major selectivity 
is provided by a multisection crystal 
filter at the input of the i-f section. The 
stages that follow will have individual 
bandwidths which are much greater 
than that of the preceding filter. 
Assume that the output of the i-f 
amplifier was applied to a product 
detector which was followed by an 
audio amplifier with a bandwidth of 4 
kHz. Since both of the noise sidebands 
present in the i-f amplifier will be 
processed by the detector, the effective 
noise bandwidth of the i-f is 8 kHz. All 
of the noise generated in the i-f ampli- 
fier (within this 8-kHz bandwidth) will 
appear at the audio output of the 
receiver. 

On the other hand, if the main 
crystal filter had a 500-Hz bandwidth, 
the only information arriving, be it 
signals, antenna noise, or front-end 
noise, will be confined to this much 
narrower spectrum. If the receiver front 
end is designed for wide dynamic range, 
the net front-end gain may be only a 
few dB. Tlius, the overall noise response 
of the receiver would be dominated by 
the noise generated in the 8-kHz effec- 
tive width of the i-f amplifier. 

There are two ways to minimize the 



Fig. 39 - A single stage of i-f amplification, 
utilizing a bipolar transistor. 


i-f noise appearing at the detector. One 
is to keep the noise figure of the i-f 
amplifier reasonably low. This is a 
partial solution. The main need is to 
restrict the bandwidth of the noise 
reaching the audio output. This means 
that additional selectivity is required 
somewhere in the receiver. 

A partial solution would be the 
addition of an audio filter within the 
audio amplifier. If this filter had a 
500-Hz bandwidth (matching that of 
the crystal filter in the beginning of the 
i-f system), the effective noise band- 
width of the i-f would be 1 kHz. The 
factor of 2 again results: Both noise 
sidebands of i-f noise are detected while 
only one contains useful information. 

The ultimate solution is to use 
proper i-f selectivity just preceding the 
product detector. If a high frequency is 
chosen for the amplifier, such as 9 MHz, 
die only useful approach is to use an 
additional crystal filter. An LC tuned 
circuit will not add enough selectivity to 
change the overall bandwidth. The filter 
in this position need not be as exotic as 
tiiat used "up front." A filter with one 
or two crystals is sufficient. 

A second approach is the use of 
multiple conversion. The signal from a 
9-MHz i-f crystal filter might be ampli- 
fied by a low-noise amplifier, tiien 
applied to a second mixer with an 
output of 50 kHz. The rest of the gain is 
obtained at this frequency, and an LC 
filter is used at the system output to 
maintain die bandwidtii the same as 
that of the original crystal filter. The 
best means for building narrow i-f filters 
in the 50-kllz region is probably to use 
ferrite pot cores. The major signal selec- 
tivity is still obtained best with the 
initial crystal filter. 

If a multiplicity of crystal filters is 
used without double conversion, the 
two filters should be well matched in 
frequency. Some filter suppliers will 



Fig. 38 — Block diagram of a modern superheterodyne receiver. 


Receiver Design Basics 87 









Fig. 40 - Gain as a function of current. 


provide matched sets of filters for a 
nominal charge. If multiple conversion 
is employed, the system is more com- 
plicated. However, the additional advan- 
tage gained is that effective decoupling 
and shielding are much easier to achieve 
at the lower frequencies. This may be an 
asset when the noise-modulation effects 
from the BFO are considered. This 
phenomenon was outlined in the section 
on product detectors. 

When choosing devices for the active 
stages in an i-f amplifier, there are a 
number of points to consider. Men- 
tioned above were overall gain and the 
ability to easily change the gain over a 
wide range. Additional problems are 
presented to the first stage. This ampli- 
fier follows the main crystal filter, 
directly. Hence, it should have an 
appropriate input impedance to termi- 
nate the filter properly. Also, this stage 
should have a low noise figure. 

Bipolar Amplifiers 

Bipolar transistors have been used 
traditionally in the i-f sections of solid- 
state receivers. If designed properly they 


Fig. 41 — A two-stage amplifier which uses 
forward and reverse age. 



may provide excellent performance 
Shown in Fig. 39 is an example of such 
an amplifier. The gain is highly depen- 
dent upon the transistor chosen. Values 
of up to 30 dB are not uncommon. If 
the amplifier is used to follow a crystal 
filter directly, it should be designed to 
have a constant, well-defined input 
impedance. This is realized through 
proper biasing of tire stage and by the 
application of feedback. The fundamen- 
tal details of the application of emitter- 
degeneration feedback were presented 
in chapter 2 in the discussion of Class A 
buffer amplifiers for transmitter applica- 
tions. Additional information on the use 
of shunt feedback is presented in the 
later discussion of ssb amplifiers. 

Depending upon the transistor used, 
there are two ways that the gain of a 
bipolar transistor amplifier may be 
changed. The more common one is the 
application of reverse age (automatic 
gain control). This is realized by 
decreasing the current flowing in the 
amplifier. The decrease in current leads 
to a decrease in the gain of the ampli- 
fier. This technique will work with 
almost any transistor that might be 
used. 

The use of reverse age in an amplifier 
has some disadvantages. First, as the 
current decreases, the input impedance 
of the amplifier will increase. This can 
cause the selectivity characteristics of 
the receiver to change dramatically if 
the amplifier follows a crystal filter 
directly. Another problem relates to the 
signal-handling ability of the amplifier. 
As the signal being received becomes 
stronger, the gain of the amplifier is 
reduced. However, as the current in the 
stage is dropped in order to reduce the 
gain, the ability of tire amplifier to 
handle the signal without distortion is 
impaired severely. 


The signal-handling ability problem 
may be circumvented by the application 
of forward age. Special transistors are 
required for such operation. However, 
since these methods are used commonly 
for i-f amplifiers in TV receivers, the 
transistors are available and inexpensive. 

Foiward age implies that as the 
current in a stage is increased, the gain 
decreases. A curve of gain as a function 
of current is shown in Fig. 40. The 
advantage of forward age is that the 
transistor is operating with tire highest 
currents when it is asked to amplify the 
largest signals. This tends to diminish 
distortion effects. Examples of 
forward-age transistors are the Motorola 
M PS-1130, MPS-H32, MPS-H01, and 
MPS6568. A number of similar devices 
are available from Fairchild Semi- 
conductor. 

Negative feedback should not be 
applied to a bipolar amplifier that is 
used for gain control. The effect of 
negative feedback is to make the stage 
gain relatively independent of the tran- 
sistor characteristics. This is opposite 
the effect desired. 

Shown in Fig. 41 is a circuit of a 
two-stage bipolar amplifier which uti- 
lizes both reverse and forward age. The 
dc biasing feedback is such that as 
current is pulled out of the age point, 
the current in the first stage will de- 
crease while that in the second will 
increase. The second stage uses a transis- 
tor chosen specifically for good 
forward-age characteristics. This ampli- 
fier has a total gain of about 50 dB, and 
exhibits a gain-control range of 80 dB. 

Most i-f amplifier devices will show 
an increase in noise figure as the gain is 
reduced. This can have the effect of 
placing an upper limit on the output 
signal-to-noise ratio of a receiver. This is 
rarely of significance in amateur 


Fig. 42 — A dual-gate MOSFET i-f amplifier. 


AGC VOLTAGE 

(APPROX . 4 VOLTS 
FOR MAX. GAIN) 



88 Chapter 5 








Fig. 43 — A differential pair i-f amplifier. 


receivers. The main objective is to en- 
sure that the noise figure does not 
increase faster than the gain decreases, 
as age is applied. 

MOSFET I-F Amplifiers 

A popular i-f amplifier is the dual- 
gate MOSFET. This device has some 
attributes that make it attractive. First, 
very high stable gain can be realized. 
The noise figure can also be made 
exceptionally low. Techniques for 
achieving low noise figures with 
MOSFETs are discussed in the following 
chapter. Finally, by changing the bias 
on gate 2 of the device, considerable 
gain reduction can be realized. An i-f 
amplifier using a dual-gate MOSFET is 
shown in Fig. 42. 

An advantage of a MOSFET ampli- 
fier is that the input impedance is 
relatively independent of the gain and 
current in the device. Furthermore, the 
distortion properties are relatively good, 



Fig. 44 - Illustration of a CA3028A differ- 
ential i-f amplifier. 


considering the low currents required. 
In spite of these assets, the device is not 
a panacea. One problem is that the noise 
figure of the MOSFET increases fast as 
the gain is decreased. Also, the distor- 
tion properties degrade markedly as 
reverse age is applied to gate 2. This is 
evident if gate 2 is made more negative 
than tire source. The reader is referred 
to the appendix for information on the 
analytical design of MOSFET amplifiers. 

IC 1-F Amplifiers 

Shown in Fig. 43 is the circuit of an 
amplifier using a differential pair of 
bipolar transistors. Although it may not 
be obvious, the two transistors are 
operating essentially in push-pull. This 
can be seen by considering the effect of 
a positive-going signal at the base of 01. 
This voltage causes the current in Q1 to 
increase. However, the emitter resistor 
common to the two stages supplies 
virtually a constant current to the pair 
of transistors. Hence, as the current in 
01 increases, that in Q2 decreases. 
Signal currents How in both transistors 
with opposite phase. 

The differential amplifier has its 
input impedance higher by a factor of 2 
as contrasted to a single-stage amplifier. 
This can be used to advantage in termi- 
nating crystal filters. 

The gain in a differential amplifier 
may often be lowered by decreasing the 
current supplied to the two emitters. 
While this could be achieved by lifting 
the grounded end of the emitter resistor 
and applying a positive potential, it is 
done more easily with an additional 
transistor. 

This brings us to a popular IC i-f 
amplifier using the RCA CA3028A. A 
circuit is shown in Fig. 44. Q3 in this 
amplifier acts as a constant-current 
source to supply the emitters. Because 
of controlled techniques applied in the 
manufacturing of ICs, Q1 and Q2 are 
virtually identical. This results in good 
balance in the outputs. Also, since the 
resistors for biasing Q3 are built into the 
IC, circuit simplification is realized. 

Reverse age is applied to the 
CA3028A by decreasing the voltage on 
pin 7 of the chip. This causes the 
current in Q3 to decrease. Since the 
collector current of Q3 is equal to the 
total current in the other two transis- 
tors, their combined gain decreases. 
While the problems outlined for reverse 
age are found in the CA3028A, the 
simplicity of the circuit makes this chip 
popular. 

Fig. 45 illustrates a cascode amplifier 
using the CA3028A. The circuit has 
some interesting properties. The input 
signal is applied to the base of Q3, and 
Q1 functions as a common-base ampli- 
fier. Because the emitter voltage of Q1 
remains fairly constant because of the 
bypass capacitor on the base of Q1 , the 


collector voltage of the input stage, Q3, 
also remains constant. This results in 
minimal capacitive feedback in the in- 
put stage, ensuring good stability and 
excellent input to output isolation. 

Under normal bias conditions, with 
no age voltage applied to the circuit of 
Fig. 46, the output current of Q3 will 
be routed directly into the emitter of 
01 . However, as current is injected into 
the base of Q2, this transistor will begin 
to conduct. As a result, part of the 
collector current in Q3 will be routed 
through Q2, causing a decrease in the 
signal flowing in Ql, the output. With 
this type of age the operating biases on 
Q3 remain constant. Because of this, the 
input impedance of the circuit remains 
constant. 

While the age range available from a 
cascode amplifier of the type shown in 
Fig. 45 is limited, the technique can be 
applied in more complicated circuits. 
An 1C i-f amplifier that uses this 
“current-robbing” method for age is the 
Motorola MC1590G. A less expensive 
cousin is the MC1350P. A circuit using 
this IC is shown in Fig. 46. 

The main advantages of the 
MC1350P amplifier come from its 
sophistication. Three differential ampli- 
fiers are contained in one package. The 
middle differential pair of transistors is 
paralleled with an extra pair that serves 
the role of current robbing from the 
main signal path. The MC1350P is 
capable of gains up to 65 dB and has 
age ranges of comparable value. A curve 
of gain reduction versus applied age 
voltage is much smoother than that of 
the typical differential amplifier. This 



Fig. 45 - Cascode i-f amplifier using a 
CA3028A IC. 


Receiver Design Basics 89 







Fig. 46 - l-f amplifier in which an MC1350P 
1C is used. 


has the effect of providing a better 
dynamic response in an age system. 

PIN Diodes in I-F Amplifiers 

Much of the discussion has been 
about the age characteristics of i-f 
amplifiers. Because this is an important 
function in the i-f system, other param- 
eters are often compromised. These 
include noise figure, linearity, and 
impedance matching. Many of these 
deficiencies may be overcome through 
the application of PIN diodes. 

The usual junction switching diode 
consists of adjacent layers of p- and 
n-doped semiconductor material. The 
junction between the two regions is 
made as small as possible in order to 
enhance the switching speed of the 
device. On the other hand, a PIN diode 
is made with a fairly large region of 
intrinsically doped semiconductor 
material between the p and n regions: 
hence the terminology of the device. 

The effect of the intrinsic layer is 
that diode action is very slow. As a 
rectifier of rf most PIN diodes arc 
nearly useless. We can take advantage of 
this. Because of the slow response time 


of the diode, it tends to behave as a 
resistor for rf currents, with the value ol 
the resistance being dependent upon the 
dc current flowing. A common relation- 
ship would be R rf = k -r /, with a typical 
value for k being around 50-ohm mA. 
Hence, if the diode is biased for 1 mA 
of dc current, the rf resistance is 50 
dims. If the dc current is increased to 2 
mA, the rf resistance drops to 25 ohms. 
The significant characteristic of PIN 
diodes is that the rf current can actually 
be much larger than the dc current. 

Hewlett-Packard is a major supplier 
of PIN diodes. Often it is possible in 
amateur applications to use high -voltage 
rectifier diodes in place of PINs, since 
the doping profile of the junction is 
similar. Sabin ( QST for July, 1970) 
recommended the Motorola MR-990A 
for this application. The diodes will 
appear resistive for rf current so long as 
the rf voltage across each junction does 
not exceed about 20-millivolts rms. The 
MR990A contains four series junctions. 

There are ways that PIN diodes can 
be used in the design of i-f amplifiers. 
Two are shown in the circuit of Fig. 47. 
In one case the diode is in series with 
the bypass capacitor in the emitter of 
the amplifier. As the dc current is 
increased through the diode, the gain 
will increase in the stage. In the other 
example the diode is in parallel with the 
collector of the amplifier. As current 
increases in the diode, the gain de- 
creases. 

If the designer is careful, he may 
construct attenuators with combina- 
tions of PIN diodes. These networks can 
have the virtue that the input imped- 
ance is fairly constant as the gain is 
varied. Such attenuators would be ideal 
in the front end of a receiver. The need 
for preserving a constant impedance 
comes from the requirement that front- 
end preselector filters need proper 
termination. Shown in Fig. 48 is an 
attenuator of the bridge-Tee variety 
which uses two PIN diodes. The pair is 
biased from a constant-current source in 
such a way that as the age voltage is 


applied the current in one diode in- 
creases as it decreases in the other. 

Switching in I-F Amplifiers 

PIN diodes are useful for switching 
functions in receivers. One application is 
for switching crystal filters in order to 
change receiver bandwidth. A related 
use would be the switching required to 
use a crystal filter for both transmit and 
receive in a single-sideband transceiver. 
A common receiver application is shown 
in Fig. 49. 

Many of the switching functions 
outlined here can be handled with high- 
speed switching diodes, like the 
1N914A. If these diodes are used, they 
should be biased so the dc current 
(lowing in them (when on) is much 
larger than the rf current being 
switched. Similarly, an “off diode 
would be reverse biased by a voltage 
which is much larger than the peak 
signal amount that will appear across it. 
If these precautions are not followed. 
IMD may occur. 

With the methods presented for 
design of i-f amplifiers, the reader may 
question which is best for his applica- 
tion. While this might be subjective, it 
will depend upon the application. For 
the typical amateur receiver where some 
IMD within the i-f amplifier is accept- 
able, the 1C approach is recommended. 
Not only is the performance adequate, 
both for gain and age capability, it is 
straightforward. 

It is interesting to note that most 
commercial equipment uses 1C i-f ampli- 
fiers. This includes receivers for the 
radio amateur as well as for TV viewers. 
On the other hand, professional equip- 
ment leans toward the use of PIN diodes 
for gain variation. Amplifiers are made 
sometimes from FETs or ICs, but are 
built also with premium-quality bipolar 
transistors. These transistors may have 
an f T in the microwave region, and are 
operated with heavy feedback in order 
to obtain stable and repeatable gain. 
The equipment described here includes 
receivers in the several-thousand-dollar 
price category, and frequency-domain 
instrumentation, such as spectrum 
analyzers. 

AGC Loops and Detection Systems 

The previous section was devoted to 
i-f amplifier design. Much of the design 
is dependent upon obtaining good gain- 
control characteristics. The gain of the 
i-f amplifier should vary smoothly with 
applied control voltage. Ideally, the 
curve of gain in dB versus applied 
control voltage should be close to a 
constant-slope straight line. The unsuit- 
able situation would be one where the 
gain change becomes large for a small 
change in control voltage. 

Fig. 50 shows a total age system. 
The main element is the variable-gain 



90 Chapter 5 




amplifier. This might be followed by a 
mixer or a product detector which 
would have a different output fre- 
quency than the one at which the main 
amplifier operates. Eventually, a low- 
impedance source is used to drive a 
diode peak detector. This produces a dc 
control voltage on capacitor Cl. This 
output is increased in a suitable dc 
amplifier and applied to the control line 
of the variable-gain amplifier. The dc 
amplifier may be inverting or nonin- 
verting, depending upon the nature of 
the desired control voltage. The choice 
is made so that an increased voltage on 
Cl leads to a decrease in gain of the 
controlled amplifier. 

There are two schemes for detection. 
One detects the i-f signal while the other 
uses the audio that is present in the 
receiver. There are valid arguments for 


each approach. While the audio-derived 
age systems are often easier to build, we 
will show that the i-f derived system is 
much better from a dynamics point of 
view. 

Consider first the case of audio peak 
detection. Shown in Fig. 51 are the 
waveforms that will result - assuming 
that initially the system is operating at 
full gain and that the age loop is 
opened at point X in Fig. 50. At some 
instant (t = 0) a strong carrier appears in 
the passband of the receiver. The 
resulting audio signal that is applied to 
tlie input of the detector is shown in 
Fig. 5 1 A. The current that will flow in 
the detector diode is shown in part B of 
the figure, while the resulting voltage on 
Cl is displayed at Fig. 5 1C. 

Consider now what will happen if 
die age system is again turned on by 





• O MAX. GAIN 


Vac V 6 MAX ATTENUATION 


Fig. 48 - Bridge-Tee attenuator using PIN diodes. 



Fig. 49 - Diode switching with PIN devices in i-f filter section of a receiver. 



Fig. 50 — Circuit representation of a total age 
loop. 


removing the open circuit at point X of 
Fig. SO. Assume that the desired maxi- 
mum level of audio output is K^pcak 
volts (Fig. 5 1 A). 

When the instantaneous voltage at 
the detector input reaches this level, Cl 
will have been charged to a level which 
will stabilize the gain. However, the 
audio cycle has just barely started. In 
reality it continues to grow, placing 
more charge into Cl. Once the peak of 
the audio cycle has been reached, no 
additional diode current flows. In all 
likelihood, the capacitor will have 
charged too far, and no additional audio 
output will occur for several cycles of 
audio output. The capacitor will slowly 
discharge through R1 until the gain 
recovers to the point where current 
pulses again flow in the diode at the 
audio peaks. Because the level is now 
changing slowly in comparison to the 
rate that the current pulses are arriving 
from the diode output, the age loop will 
now follow the strength variations of 
the arriving signal, holding the output 
fairly constant. However, the initial 
overshoot described not only causes a 
large click or thump in the receiver 
output, but may cause information to 
be lost for a short period. 

The answer to stabilization of the 
audio-derived loop is to add some resis- 
tance in series with the diode (or to 
increase impedance of the diode driver). 
This will slow the response to the point 
that the capacitor Cl may not became 
completely charged by one cycle of 
audio. Unfortunately, this reduces the 
rate that i-f gain is reduced and leads to 
the initial information causing excessive 
receiver output. 

Consider now the case of an i-f 
derived detection system. This is shown 
in the set of curves shown in Fig. 52 
where the time scale is essentially the 
same as that used for the audio-derived 
case. There are a number of different 
features. First, the rate that current 
pulses from the diode detector are 
applied to the memory capacitor, Cl, is 


Receiver Design Basics 


91 






Fig. 51 — Waveforms for open-loop audio age 
detector. 


much higher. Hence, the impedances 
may be adjusted so that a single pulse 
does not charge the capacitor com- 
pletely, without seriously slowing down 
the loop response time the cause of 
overshoot effects. Second, even though 
tlie signals arriving at the input to the i-f 
filter of the receiver may all be constant 
in amplitude, the resulting filter output 
will not reach a stable amplitude 
immediately. This is because any filler 
has a rise time that is related to the 
filter bandwidth. The narrower the fil- 
ter, the longer the rise time will be. In 
this situation, the age loop is capable of 
responding fast enough that the gain 
will adjust itself so that the input signal 
is followed. The bandwidth of the con- 
trol system should be wide in compari- 
son with that of the information 
arriving from the i-f filter. 

In spite of the deficiencies of audio- 
derived age detectors, they may be used 
with satisfactory results in some 
receivers. The transient problems out- 
lined here are much more severe when 
designing a cw receiver than they are 
with ssb. This is especially true if the 
BFO is adjusted to provide a low-pitch 
beat note. 

A few tricks may be applied to 
improve the attack characteristics of 
audio-derived systems. The first is to 
employ full-wave detection instead of 
the half-wave type outlined in Fig. 51. 
Full-wave detection may be achieved 
with a center-tapped transformer, or 
with a pair of op amps. Examples are 
shown in Fig. 53. 

Another method is the judicious 
application of clipping. A sample circuit 
is shown in Fig. 54. In this case the 
response time of the loop is slowed by 


addition of resistance in series with the 
diode detector. However, the audio 
output of the receiver is prevented from 
becoming excessive (thus protecting the 
operator's ears) by limiting the level of 
audio signal applied to the receiver 
output and to the age detector. The 
control in Fig. 54 should be adjusted so 
the clipped peak voltage at the detector 
is about 3 dB above the level that the 
age loop establishes eventually. If an 
oscilloscope with good triggering charac- 
teristics is available, the dynamics may 
be adjusted so stabilization will occur 
within about ten cycles of audio output. 

Shown in Fig. 55 is a pair of age 
systems that may be applied with i-f 
amplifiers using CA3028A or MC1350P 
ICs. These circuits may be used with 
audio or i-f detection. In each case a 
JFET is used as the input to the error 
amplifier. Suitable npn transistors are 
2N3565s, 2N2222As or any equivalent 
silicon device. The pnp transistors are 
similarly uncritical. Good choices would 
be the 2N3906 or the 2N3638. The 
controls shown in the error amplifier 
(R2 and R3) should be adjusted for the 
proper voltages during full-gain condi- 
tions. These voltages are marked in the 
schematics. The systems also include 
means for manual control of the gain. 

The FET type is arbitrary. Almost 
any FET will work, since it is used in a 
circuit with heavy feedback. The pinch- 
off should not be more titan 5 or 6 
volts, but other parameters are not 
critical if the supply voltage is 12 or 
more. In each circuit provision is made 
for muting the amplifiers. That is, by 
grounding the point marked “M" the 
gain of die i-f may be reduced to its 
minimum value. 



Fig. 52 - Characteristics of an i-f derived age 
detection system. See text. 


In the two systems of Fig. 55 the 
recovery time is determined by the time 
constant, r = R1 Cl. For the longer 
recovery times desired for ssb. the time 
constant should be I to 2 seconds. One 
deficiency of these circuits is that the 
stronger signals will cause Cl to charge 
to a slightly higher voltage. Because of 
this, the time will then be somewhat 
longer for full gain to return. 

Fig. 56 shows an agc-detection sys- 
tem that overcomes this deficiency. This 
circuit may be used with i-f or audio- 
derived detection. A pair of detectors is 
utilized to produce a full "hang" action. 



92 Chapter 5 










Fig, 54 - Audio limiter for use with af types 
of age loops. 


Diode CR1 serves as the main age 
detector, with the following amplifier 
being adjusted to drive MC1350P or 
MC1590G amplifiers. The system could 
be adapted for the reverse age of the 
CA3028A, or for virtually any i-f 
characteristic. 


The action of the two loops is 
explained by considering sequentially 
how the circuit behaves. First, consider 
the effect of a short pulse of noise. This 
pulse will produce a lengthened 
response at the output of the i-f filter, 
which is detected ultimately by CR1 to 
cause a momentary reduction of tire i-f 
gain. Audio output will result in the 
receiver and will also cause a signal to 
appear at CR2 and CR3. Because of the 
100-kfi resistor in series with CR2, C2 
will acquire a small charge from this 
pulse. As a result, the main memory 
capacitor, Cl, will discharge quickly 
through R3 and the drain of Q3. 

On the other hand, consider the 
effect of a carrier, a string of cw 
characters, or a ssb signal. CR1 will 
again charge Cl , and will lead to a gain 
reduction in tire i-f system. The sus- 
tained audio signal that results will 
cause CR2 and CR3 to operate and 
charge capacitor C2 negatively. The gain 
of the op amp driving these diodes is 
adjusted so that normal signals cause the 
op-amp output to swing over its full 


range, from ground to the positive 
supply. This will cause C2 to be charged 
to a high negative voltage. The value will 
be approximately twice the supply volt- 
age at U2. In this condition Q3 is 
pinched off. Because of this the only 
discharge path for the main memory 
capacitor, Cl, is through Rl, a 22- 
megohm resistor. When the signal 
disappears, C2 begins to discharge 
through R2. When the voltage at tire 
gate of Q3 becomes close to ground, so 
the FET is no longer in a pinch-off 
condition, Cl is discharged quickly 
through Q3. 

Listening to a system of this kind is 
enlightening after being accustomed to 
the simpler methods. With the full hang 
age, the receiver is virtually silent after a 
strong signal disappears from the pass- 
band. However, after a timing period 
associated with the C2-R2 timing net- 
work, the receiver returns to full gain 
within roughly 50 milliseconds. The 
time delay is virtually independent of 
the strength of the incoming signal. 

An audio signal is suitable for driving 



Receiver Design Basics 93 




r 



secondary detector, CR2, because a 
slow response is desired in this loop. An 
i-f derived signal could be used also. The 
741 op amp, U2, would need to be 
replaced with a circuit suitable for the 
i-f frequency used. 

An age system of this kind is used in 
a receiver at W7ZOI. It will be described 
later. The age characteristics have been 
studied extensively by means of a trig- 
gered oscilloscope. No sign of overshoot 
or pumping could be detected with 
signals ranging from the minimum 
detectable amount up to 50 mW at the 
antenna terminals. Higher levels would 
probably endanger the front-end com- 
ponents of the receiver. The signal to be 
detected was derived from a 9-MHz i-f 
amplifier. 

With the age systems outlined, addi- 
tional gain may be required in order to 


drive the detector diode. This extra gain 
is usually minor with audio-derived 
systems, since the levels are already high 
when that part of the receiver is 
reached. With an i-f derived detector, 10 
to 40 dB of additional gain is often 
required, depending upon the overall i-f 
gain. Care should be taken to ensure 
that the age detector is not activated by 
die BFO energy. BFO energy should be 
confined to the product detector, as 
outlined earlier. 

The age threshold of a receiver (the 
level at the antenna terminal where age 
action begins) is determined by the 
characteristics of the detector diode and 
die gain ahead of the detector. For most 
applications a suitable threshold is —100 
to 1 10 dBm. 

The “tightness" of an age loop can 
be expressed in a number of ways. 


Usually, the variation in audio output in 
dB is given for an input variation from a 
few dB above threshold to a level 60 or 
80 dB stronger. This figure of merit, no 
matter how it is defined, will depend 
mainly on the overall fixed gain in the 
age loop and upon the age character- 
istics of the i-f amplifier. 

Simple Superheterodyne 
Front-End Design 

Of all of the parts in a receiver die 
front end is probably the most critical. 
A poor design can lead to disastrous 
results. A proper design will yield 
acceptable performance. This receiver 
section is so critical that we have 
devoted an entire chapter to its design. 
Special attention is paid to the problems 
of noise figure and dynamic range. The 
criterion for optimizing either is pre- 
sented with a discussion of the tradeoffs 
between the two. 

While not difficult, die subject of 
front-end design is complicated enough 
diat it cannot be approached casually. 
In til is section some information is 
presented for the beginning experi- 
menter. Totally acceptable performance 
for general-purpose applications may be 
attained if a few precautions are fol- 
lowed. Some sample circuits are given 
with rules of thumb for their use. The 
reader is referred to chapter 6 and to 
the appendix for design details. 

Block Diagram 

The front-end section of a receiver is 
diat portion containing the first mixer, 
preselection filters and perhaps an rf 



Fig. 57 - Block diagrams of receiver front end for single-conversion circuits. 


94 Chapter 5 





Fig. 58 - Circuit for a dual-gate MOSFET mixer. 


amplifier. The standards drat must be Furthermore, it is wise to protect the 
met arc to provide sufficient receiver receiver from signals other titan those to 
noise figure and image rejection. Gain is which the receiver is tuned. In many 
often desired, although not always receivers this front-end selectivity is 
necessary. Shown in Fig. 57 are block provided with a single or a double-tuned 
diagrams for the front end of single- circuit. The latter is preferred, owing to 
conversion receivers. the improved skirt selectivity for a given 

The two systems differ only in the 3-dB bandwidth. The design of simple 
inclusion of an rf amplifier in the preselector filters is covered in some of 
second. The first contains none. Both the sample circuits. The subject of 
circuits have a preselector network and loaded and unloaded Q was covered in 
a mixer. The most tragic mistake made chapter!, 
by the beginning experimenter is that he 
uses an rf amplifier when it is not really *' lxer Circuits 

needed. The only purpose of an rf There are a number of semi- 
amplifier in a receiver front end is to conductors that will function well as 
reduce the overall noise figure. This will mixers. Of all that are available the 
enhance the sensitivity of the receiver, simplest to use is the dual-gate 
However, on most of the lower fre- MOSFET. A circuit is shown in Fig. 58. 
quency amateur bands an acceptable A single tuned circuit is used as the 
noise figure may be obtained with a preselector. A tuned transformer at the 
mixer front end. The effect of the rf output matches the crystal filter that 
amplifier is to increase the signal levels follows the mixer, 
at the mixer, causing a degradation in The gain realized with this circuit 
signal -hand ling ability. will depend upon exact device param- 

A standard for evaluating a receiver elers. Values of 15 dB are representa- 
lor sufficiently low noise figure was tive. The proper LO injection level for 
presented at the beginning ol this this mixer is 5 volts pk-pk. Lower levels 
chapter. It bears repeating: When the will decrease gain and will compromise 
antenna is connected to the receiver, the dynamic range. The noise figure of this 
output noise should increase signifi- front end is often 8 to 10 dB. This is 
cantly. II this criterion is met there is no low enough to ensure usable sensitivity 
need to seek a lower noise figure, in alnvst all hf applications. 

Generally speaking, the atmospheric and The dual-gate MOSFET appears to 
man-made noise levels Irom 1.8 to 21 present a very high impedance at its 

MHz are high enough that an rf ampli- input (gate 1) in the hf region. Because 
tier is redundant. of this, the tuned circuit is singly 

Image rejection must be maintained, loaded. The loaded Q of the preselector 


Fig. 60 — Example of capacitive-divider 

Fig. 59 - Method for capacitive matching at matching to decrease the impedance level at 
the input of a receiver. the gate of a MOSFET. 



is determined by the unloaded-0 value 
of the inductor and the loading pre- 
sented by the 50-ohm antenna. 

The values shown in Fig. 58 are for 
an input on the 20-meter band. The 
inductor hasa(7„ of approximately 200 
and consists of 20 turns on a toroidal 
form. The antenna link contains 2 turns. 
Because impedances transform 
according to the square of tire turns 
ratio with toroidal cores, the equivalent 
resistance across the coil is 5000 ohms. 
The inductance is nominally 1.5 /iH. 
The equivalent parallel resistance repre- 
senting the unloaded Q is of the order 
of 27 kf2. Since this value is large when 
compared to the 5000 ohms repre- 
senting the antenna loading, the losses 
in the circuit will be small, The loaded 
Q will be 5000 (2nfL) = 37.4. (See 
chapter 2 for details.) The 3-dB band- 
width of this circuit will be 14,000/37.4 
= 374 kHz. No tuning would be re- 
quired for the complete 20-meter band. 
It would be needed for the lower bands. 

If a higher loaded Q was desired in 
the preselector, it could be obtained by 
changing the turns ratio. For example, 
the link could be reduced to a single 
turn. This would produce a Q L value of 
85. The value might be higher. This is 
because with only 1 turn for the anten- 
na link, the coupling may become weak 
enough that the turns squared relation- 
ship no longer applies. A loaded Q of 85 
would imply a bandwidth of 165 kHz. 
It may be shown that the insertion loss 
of the filter will now be much higher 
(nearly 10 dB), which would degrade 
noise figure. This is not desired. 

An additional problem with the 
liigher turns ratio configuration is the 
higher signal voltage appearing at the 
input of the MOSFET. This could com- 
promise dynamic range. A lower voltage 
at the input may be realized by tapping 
the gate down on the tuned circuit. This 
will not alter tire loaded Q of the 
preselector, nor will it reduce insertion 
loss. The tap may be on the coil, or it 
may be composed of tapped capacitors. 

The method of capacitive matching 
is shown in Fig. 59 where it is applied to 
matching of the antenna. If the antenna 
resistance is R a (usually 50 ohms) and 
the equivalent resistance presented 


Fig. 61 — Method for using a single series 
capacitor at the receiver input to match a low- 
impedance antenna system to the input stage. 



Receiver Design Basics 95 






Fig. 63 - A singly terminated double-tuned input circuit. 



Fig. 64 - Bipolar-transistor mixer with LO-energy injected at the emitter. 


ANT. | 


s # 


RF AMPLIFIER 
too E-SOO 22 



Fig. 65 - Circuit of a common-gate JFET rf amplifier. 


across the coil is R c , the two are related 
with 



Rc = 


Ro 

" 4 8 )5 


(Eq. 6) 


Using this equation, it may be shown 
that a 9:1 capacitance ratio would 
produce the same 100:1 impedance 
transformation that the link on the coil 
of Fig. 58 afforded. 

If a capacitive transformation is used 
to decrease the impedance level driving 
the gate of the MOSFET (Fig. 60), care 
should be used. A resistor would be 
required from the gate to ground to 
establish a proper dc bias. This resistor 
should be very large in ohmic value. 
Otherwise, it might load the coil exces- 
sively. In a single tuned circuit, the 
loading should come from the antenna 
and not from extra resistors that are 
added. 

A third method for matching into 
the resonator would be to use a low- 
value capacitor directly between the 
antenna terminal and the “hot” end of 
the tuned circuit. This is shown in Fig. 
61. The equations for applying this 
method are examined in the appendix in 
connection with the filter tables. 

In the mixer circuit of Fig. 58, a 
tuned transformer was used to match 
between the drain of the MOSFET and 
the crystal filter that follows. With 
almost all MOSFETs that are used in 
mixer applications, the output imped- 
ance is very high. Values of 100 k£2 or 
more are representative. If the trans- 
former were designed to match between 
this level and the 500-ohm input to a 
filter (symbolic of the KVG line of 
9-MHz crystal filters), the dynamic 
range of the mixer would be compro- 
mised severely. It is mandatory that a 
resistance be placed across the coil. This 
ohmic unit establishes a well-defined 
termination for the filter and limits the 
impedance presented to the drain of the 
mixer. 

In the circuit of Fig. 58, the drain 
transformer has a 30:7 turns ratio. This 
causes the 1 0-kS2 resistor to appear as a 
500-ohm termination for the filter. An 
equally viable (and often desirable) cir- 
cuit for the output would be a pi 
network. It should be designed for a Q 
of 10. 

The single tuned circuits that have 
been used for preselection are often 
lacking in skirt selectivity. This will 
compromise image rejection. A better 
circuit is a double or triple tuned one. 
Shown in Fig. 62 is a double-tuned 
front end. Again, only a mixer is used. 
No constants are given, since they will 
depend upon the band of interest. 
Specific designs are presented in the 
filter tables of the appendix. 


96 Chapter 5 











Fig. 66 - A bipolar-transistor rf amplifier 


A resistor is shown at the output of 

unc P tf S T leCt0rS ' from the 8 ale of the 
MOSf-fc r mixer to ground. This resistor 
is necessary to terminate the filter prop- 
erly. These filters are classed "doubly 
terminated,’’ and are representative of 
the tillers in the appendix. It is not 
necessary that double-tuned circuits be 
doubly terminated. Suitable circuits 
may be realized with antenna loading as 
the only termination. See Fig. 63. This 
will alter the designs from those given in 
the appendix. The best approach for 
using such filters is empirical. The 
coupling capacitor (C3) should be vari- 


able. Initially, it should be adjusted for 
minimum capacitance. The resonators 
are then peaked (Cl, C2). The input is 
swept to ensure that a single response is 
provided. Then, coupling capacitor C3 is 
increased slightly, and Cl and C2 are 
peaked again. 

This procedure is repeated until a 
double-humped type of response ap- 
pears The coupling-capacitor value is 
then decreased slightly and left in that 
way. II the bandwidth obtained with 
tins course is too narrow, the loading at 
the antenna terminal may be increased 
(more turns on the link). The process is 


-* »ivnj uuiiuwiuill 15 

obtained. Tire builder should use the 
filters in the appendix as a guideline for 
the approximate values to begin with in 
his (or her) empirical realization of a 
singly terminated filter. It is not recom- 
mended that three (or more) filter 
sections be attempted unless each end 
of the filter is terminated properly. 

While we have strongly recom- 
mended the dual-gate MOSFET mixer 
there are other devices that will perform 
suitably for such applications. These 
include many ICs which were discussed 
m the product-detector section. Bipolar 
transistors will also perform as mixers. 
A typical circuit is shown in Fig. 64. 
the LO is injected onto the emitter of 
(lie mixer. Best performance will be 
obtained from this circuit if large dc 
bias currents are used. Bipolar mixers 
are not recommended. 

Some of the ICs that are used as 
mixers are the MC1496G and 
CAj 02,XA. They have the advantage of 
balance. This reduces the amount of LO 
power that might appear at the antenna 
terminal. These devices are usually more 
inject to overload effects than the 
MUM' LI is. A receiver described later 

JV! 1 A S ,n hapter shows an a PPhcation of a 
t A3 02 8 A mixer. 

RF Amplifiers 

It is sometimes desirable to use an rf 
amplifier ahead of a mixer. Special 
applications where inclusion could be 




*0 

tLlOOpF 

x ™ 

- -AAAKIt! 

^»ooo 

< 

1 < 

1 

>2200 

> 4700 

S AMP 

0? 

i QAiF 

-♦|/ n 

1 2N3S63 

Cr 

It ° 

\ r 

\ 01 

\2NS365 

“7 

) '-i 

X iv > 

2200 


At 

OUT 


EXCEPT AS INDICATED, DECIMAL 
VALUES Of CAPACITANCE ARE 
IN MICROFARAOSCjiFi; OTHERS 
ARE IN PICOFARADS ( pf OR „j,F ) 
RESISTANCES ARE IN ohms; 

» *1000 . M • 1 000 000 


***• **• - g; *»»• 

“ - « » •"»». »l» on LI. SI - SSSSTmcwoT 


Receiver Design Basics 


97 









Front panel of the direct -conversion receiver 


desirable would be for 10-meter or vhf 
reception. Alternatively, they might be 
included in the front end of portable 
receivers to be used in isolated locations 
which are devoid of man-made noise. 
These locations do exist. 

Shown in Fig. 65 is the circuit of a 
simple rf amplifier that is recommended 
for general-purpose applications. A 
JFET is employed in the common-gate 
configuration. This circuit will provide a 
gain of 8 to 14 dB, depending upon the 
FET characteristics. The input imped- 
ance at the source will be low. Repre- 
sentative, values are from 100 to 300 
ohms. The output should be a tuned 
circuit with a high L-C ratio. This 
maximizes the impedance presented to 
the drain, increasing the gain. The resis- 
tor in the drain suppresses uhf, vhf and 
parasitic oscillations. The general im- 
pression that common-gate FET ampli- 
fiers are unconditionally stable is not 
true. 

Shown in Fig. 66 is a circuit for a 
bipolar transistor rf amplifier. A com- 
mon thought among amateurs is that 
bipolar transistors are not suitable for 
front-end applications because of over- 
load. This is not absolutely true. If 
low -noise transistors with high values of 
f T are used in circuits with negative 
feedback, excellent performance may be 
obtained. The circuit shown is not 
subject to easy overloading. This results 
from the feedback and high bias current 
(20 mA). The input and output imped- 
ances are both close to 50 ohms. This 
makes the circuit easily matched to 
filters from the appendix. Bipolar tran- 
sistors are not recommended unless 
these precautions arc heeded. 

The amplifier of Fig. 66 has a gain of 
nearly 20 dB. The noise figure is not 
low, but is reasonable. One represen- 
tative sample investigated showed a 
6.5-dB value. The bandwidth is over 1 00 
MHz, making the circuit useful for all hf 
bands. The extensive feedback does 
ensure stability. 

A Two-Band Direct-Conversion 
Receiver 

There is often a need for a simple 
receiver which still offers good per- 


formance. An example would be a 
compact receiver for portable or emer- 
gency operation. Another might be a 
club project where a number of begin- 
ners build a “first station.” The receiver 
described in this section is aimed at 
these applications. 

The detector and audio circuit is 
shown in Fig. 67. An MC1496G IC is 
used as the product detector. Ample 
audio gain is provided by a pair of 
transistors. In the interest of simplicity, 
minimum audio selectivity is used in the 
system. However, an R-C active filter 
could be added at the audio output, if 
desired. 

The detector differs from that nor- 
mally used with this 1C. First, the gain is 
increased significantly by placing a 
bypass capacitor between pins 2 and 3 
of the chip. The more typical applica- 
tion is with a resistor (100 to 1000 
ohms) in this position. The other de- 
parture from the standard circuit con- 
cerns the bias current used. This is 
determined by the resistor connected 
between pin 5 and the positive supply. 
The usual 10-kf2 resistor has been re- 
placed by a 3300-ohm one. This in- 
creases the gain and signal-handling ca- 
pability of the detector by about 10 dB. 

The input circuit will tune from 
approximately 3 to 8 MHz. This allows 
the 80- and 40-meter amateur bands to 
be tuned without band switching the 
front end. Other tuned circuits may be 


substituted in order to cover additional 
frequencies. A 10-pF capacitor is used 
between tire tuned circuit and pin 1 of 
the 1C. For operation on the 160-meter 
band, a suitable value would be 22 pF. 
For operation at 14 or 21 MHz, the 
value should be decreased to 5 pF. 

A wide range oscillator is shown in 
Fig. 68. A JFET is employed in a 
Hartley circuit. A buffer/amplifier with 
two bipolar transistors is used to obtain 
ample BFO drive voltage. A 3/8-inch 
diameter slug-tuned coil is used with 
parallel capacitors (air variable and cer- 
amic NP0) to form the resonator. With 
the band switch (an inexpensive toggle- 
type) open, the oscillator tunes from 6 
to 8 MHz. When tire switch is closed, a 
360-pF silver-mica capacitor is paral- 
leled with the others, providing a tuning 
range of 350 kHz in the 80-nreter band. 
The exact range desired may be ob- 
tained by adjustment of the coil slug. 

An experiment was performed to 
move the oscillator higher in frequency. 
The slug was removed from the coil and 
all fixed-value capacitors were discon- 
nected. In this condition, the oscillator 
would tune to about 15 MHz. The 
stability was adequate for reception of 
cw and ssb signals. 

A pc layout is shown in Fig. 69 for 
the detector and audio board. The size 
is approximately 2X4 inches. The 
experienced builder may wish to minia- 
turize the circuit further. But. the begin- 



Inside view of the direct-conversion receiver. The antenna trimmer is at the left. Seen at the 
bottom of the box is the oscillator board. A strip of flashing copper serves as a ground bus. 


98 Chapter 5 



ner may find it desirable to expand the 
size especially if small components are 
not available. The existing layout will be 
cramped unless rather small 0.1 -^F ca- 
pacitors are used. 

The VFO is built on a 3 X 3 inch 
piece of unclad circuit board with rivet- 
in terminals for solder connections. (A 
board could be etched for this circuit.) 

The two-band receiver is packaged in 
a 2 X 4 X 6 inch chassis. No vernier 
drive mechanism was used. Instead, two 
tuning capacitors are used in parallel. 
One functions as the main tuning while 
die other serves as a bandspread control. 
The advantage is one of mechanical 
simplicity, allowing quick completion of 
the project. Accurate calibration is not 
easily realized with this method. 

The results obtained with this re- 
ceiver were gratifying. Unlike some pro- 
jects, this receiver functioned as de- 
signed when power was applied. Cw and 
ssb quality are excellent. 

This receiver might serve as a step 
toward construction of a simple super- 
het. After being built as shown, a crystal 
filter could be added. The VFO can be 
moved easily to any frequency in the 3- 
to 15-MHz range, as outlined earlier. 
The addition of a dual-gate MOSFET 
mixer and a crystal -con trolled BFO 
would result in a superheterodyne sys- 
tem (see Fig. 70). 

The builder might want to add an rf 
amplifier, especially if the receiver is to 
be used on one of the higher bands. A 
suitable circuit using a 2N5179 is shown 
in Fig. 71. For dynamic-range reasons, 
one might scowl at the use of a bipolar 
transistor instead of an FET. However, 
this opinion is not valid. 

The amplifier shown is broadband, 
has 50-ohm input and output imped- 
ances, and provides nearly 20 dB of 
gain. The use of heavy feedback ensures 
stability. Good signal-handling ability 
results from a high bias current (20 
mA). The input preselector networks 
are in the appendix at the end of the 
book. 

A Pocket-Size Direct-Conversion 
Receiver for 40 Meters 

Solid-state technology permits mini- 
aturization and low power con- 
sumption. The receiver of Fig. 72 was 
built to take advantage of both assets, 
while offering simplicity of construc- 
tion. 

The pocket portable uses two tran- 
sistors and two ICs. Power is provided 
by a small battery contained in the 1 X 
3-1/2 X 5-1/2-inch aluminum cabinet. 
The receiver is built on a 2-1/2 X 3-1/2 
inch double-sided pc board (one side is 
all ground foil). Only 1 1 mA of current 
are required from the 9-volt battery. 

Tire 40-meter cw band was chosen. 
Fig. 69 - Foil-side circuit board pattern and parts layout for the detector and audio circuit of The receiver could be adapted to any of 
Fig. 67. Drawing is to scale. the bands from 1 .8 through 1 4 MHz. 

Receiver Design Basics 99 


if* 




The main board for the receiver. The input tuned circuit is at the left, adjacent to the product- 
detector 1C. An audio amplifier is contained on the remainder of the board. 


BANDSPREAD 



C2 - Miniature 20-pF air variable. 

03 - 200-pF mica trimmer. 

CR1 — High-speed silicon diode, 1 N91 4A or 
equiv. 

L3 - 20 turns No. 24 enam. wire on 3/8-in. 
dia. ceramic slug-tuned form (Miller 


4400-2 form), tapped 5 turns from 
ground. 

03 - JFET, MPF102, HEP802. or TIS-88 
suitable. 

SI - Spdt miniature toggle. 

VR1 - 6.2-V, 400-mW Zener diode. 


Fig. 68 - Schematic diagram of the tunable oscillator for the receiver of Fig 67 Fixed-value 
capacitors are disk ceramic unless otherwise indicated. Fixed-value resistors are 1/2-W com- 
position. 


365 pF.VAR 


*12 



■OuF J 4 

AF OUT 


2200 








Fig. 70 — Details of how a mixer and BFO can 
be added to obtain a superheterodyne receiver 
with the circuit of Fig. 56. 



Fig 71— Suggesled r-t amplifier tor use with the universal direct-conversion receiver. 

Ins 

par 





Inside layout of the receiver. All of the circuit 
part of the local oscillator. 


The product detector uses a Motor- 
ola MFC8030 differential-amplifier, 
This IC is similar to the CA3028A, 
except that external biasing resistors are 
required. This adds to the parts count, 
but allows the IC to be biased for 
minimum current — a major design goal. 

The detector output is applied to a 
2N3906 pnp amplifier. This is routed 
through the audio-gain control to an 
MFC4010A. This tiny four-terminal IC 
is barely larger than a plastic transistor. 
It contains three direct-coupled stages. 

The VFO uses a bipolar transistor in 
a Colpilts circuit. For minimum power- 
supply current, no Zener-diode regula- 
tion is employed. A ceramic slug-tuned 
coil is used with an output link to drive 
the detector. The stability is adequate. 

In spite of simplicity the receiver 
performs well. Sensitivity is good. Sig- 
nals from four continents were heard 
(on cw) during the first evening of use. 
Selectivity is poor, but could be im- 



External view of the 7-MHz portable receiver. 
The controls are, left to right, af gain, tuning, 
and on-off switch. 


is on a single pc board. The slug-tuned coil is 


proved with audio filtering. Because 
miniature projects like this one are 
dependent upon the size of the compo- 
nents available, no pc layout pattern is 
offered. 

A Simple Superhet for 
80 and 40 Meters 

In the 1950s nearly every issue of 
the Handbook contained a receiver 
which covered 80 and 40 meters. The 
basis of the design was a superhetero- 
dyne utilizing single conversion with an 
i-f of 1.7 MHz. The oscillator tuned 
from 5 2 to 5.7 MHz. Witli this set of 
frequencies, one band was the image of 
the other. This led to simplification, 
because band changing was realized by 
tuning the front-end preselector. 

Shown in Fig. 73 is a solid-state 
version of the Handbook classic. This 
receiver was built by Jeff Danun, 
WA7MLII. 

Only eight semiconductors are used 
in the receiver. Three dual-gate 
MOSFETs serve as the input mixer, i-f 
amplifier, and product detector. The 
rest of the functions are provided by 
means of bipolar transistors. Selectivity 
is obtained with a homemade two- 
crystal filter of tire half-lattice type. 

Circuit Details 

The input mixer uses a 40673 
MOSFET with a single tuned circuit as 
the preselector. A half-wave filter is 
included in the antenna line to suppress 
spurious responses from high-order pro- 
ducts created in the mixer. The filter is 
cut for a 7-MHz center frequency. The 


low-pass nature of the filter allows 
80-meter signals to pass unattenuated. A 
short piece of coaxial cable is used to 
connect the panel-mounted variable ca- 
pacitor in the preselector to the circuit 
board. 

The drain of the mixer feeds the 
tuned primary of the transformer sec- 
tion of the crystal filter. The secondary 
is a center-tapped 12 -turn winding. To 
ensure good balance, this winding is 
wound as six bifilar turns. The crystals 
were ordered for 1700.0 and 1700.3 
MHZ. To keep the cost down, a .01- 
percent tolerance was specified. When 
the crystals arrived, their separation was 
only 200 llz. While each crystal was 
within the manufacturer's specification, 
the bandwidth was narrower than de- 
sired. If the receiver is to be used for the 
reception of ssb as well as cw, a sepa- 
ration of 1 .5 kHz is recommended. With 
the existing filter, cw selectivity is im- 
pressive. Single-sideband stations can be 
copied, but the audio sounds distorted. 

A I0-kf2 resistor is used to terminate 
the filter. This value was arrived at 
experimentally. It assured minimum 
filter loss without passband ripple. 
Other values may be required, de- 
pending on the crystal characteristics. 

A stage of i-f gain is provided by Q2, 
a dual-gate MOSFET. While the gain is 
not hijjh, it is enough to overcome the 
loss of the crystal filter. Some variation 
of i-f gain is provided with a front-panel 
switch. In normal operation, gate 2 of 
02 is biased at about 4 volts. However, 
when the switch is closed, the bias on 
gate 2 is reduced to .0. This causes a 
decrease in stage gain of approximately 
20 dB. In the unit built by WA7MLH, 
this switch is activated by pulling on the 
audio-gain control knob. The builder 
could use a separate switch. 

A tliird 40673 MOSFET, Q5, is the 
product detector. This stage is typical of 
many using a FET, except that the bias 
for gate 2 (where the BFO is injected) is 
from a grounded resistor. The typical 
circuit has this resistor returned to the 



Exterior of the 80- and 40-meter superhetero- 
dyne built by WA7MLH. The box measures 
5X6X9 inches. 


Receiver Design Basics 101 




OSCILLATOR 


BUFFER 

(OOO 100 


5 2-57 MHj 


22kS 05 

< 2N3904 


nVRA 

1 16.8 V 

7Tw- T 


.fa_ 

L 100 

-680 

r 

* 

L 7 '100 - 

1 

" S M 

' S "X 

► 




-680 < 




[~ S.M. < 

^ A 

* 3300 

L, 


04 

2N3904 


l-F ATTEN. / 3 : 


l-F AMPLIFIER 


^ji 

HALF-WAVE 

1 

FILTER 


7 MH* 


!C2 -L 470 
-180 ^-J-T m. 


FILTER I G2 / 

. n 1 . s'V 


Y1 

1700 KHj 



,00 >100 0, \ i -00 * > 


’Is > 1000 \2 


•100 < =C.05 

• 100 < 


PRODUCT 

DETECTOR 


^^wv- 

K 

s’ 000 

40673 \ 

G2 /T^£ i 

G'( lT)s 




>120 - ~ 05 ' 

rr~7 

( 

^D-7 0TO 

1 1- 


10>jF 

1000 «v 


.05 ™.05 


* 1 I 



50/.F-H 
15V 'T' 


07 

2N3565 


AUDIO AMPLIFIER 
2200 


08 

2N3565 


IL iOhF 

'5V 



200 

S.M. 

1/ . 

J 

U \ 
< 

— 1— 2200< 

r _^C4 


7'15 

/ 7 

L4 

" 

79 

— f 

„H 

—L. 2200 


T'slT- 


\VR2 

16.8V i.1 
/ 1W 


S.M . SILVER MICA 

EXCEPT AS IN0ICATE0, DECIMAL 
VALUES OF CAPACITANCE ARE 
IN MICROFARADS I JlF I OTHERS 
ARE IN PICOFARADS (RF OR jlJlFr, 
RESISTANCES ARE IN OHMS’, 

> <1000. M- 1 000 000 


Fig. 73 Schematic diagram of the 40- and 80-meter superheterodyne receiver. Fixed-value capacitors are disk ceramic unless otherwise 
noted. Fixed-value resistors are 1 /2-Watt composition. Polarized capacitors are electrolytic. 


Cl — Miniature 365-pF variable. 

C2 - 180-pF mica trimmer. 

C3 - 100-pF air variable, (See text.) 

C4 - 15-pF variable. (See text.) 

J1 - Antenna receptacle of builder's choice. 
J2 - Two-circuit phone jack. 


LI - 3 turns No. 26 enam. wire over L2. 

L2 — 36 turns No. 26 enam. wire on Amidon 
T68-2 toroid core. 

L3 Approximately 1 .57-pH. slug-tuned 
coil (Miller 42A156CBI). 

L4 - 7.9-pH slug-tuned coil (Miller 43105- 
CBI). 


R1 10,000-ohm audio-taper carbon control. 
SI - spst toggle. 

SI — spst toggle. 

T1 — Primary, 53 turns No. 28 enam. wire on 
an Amidon T68-2 toroid core; secondary, 

12 bifilar turns. 


102 Chapter 5 



1 



Interior view of the WA7MLH receiver. The mixer front end is at the left, and the crystal filter 
is at the center of the board. At the right can be seen the product detector and audio section. 


source of the 40673. The technique 
used led to a simplification. 

Audio gain for the receiver is ob- 
tained from a pair of 2N3565s. Ample 
gain is provided for ear-shattering head- 
phone output. 

Both oscillators in the receiver use 
the standard Colpitts format. The main 
L0, which covers 5.2 to 5.7 MHz, is 
tuned with a single-section capacitor 
(C3) from a surplus BC-454. Any vari- 
able capacitor with a range of at least 
100 pF will serve as well. With other 
capacitors, a vernier mechanism is 
recommended. It was not needed with 
the surplus capacitor since a high qual- 
ity gear mechanism and dial drive are 
part of the capacitor unit. 

While a commercially available coil 
was used for the LO tuned circuit, tire 
inductor in the BFO was a junk-box 
item. A suitable substitute would be a 
J. W. Miller 43105CBI. The BFO is 
tunable from the front panel by means 
of a 1 5-pF variable capacitor. 



Local-oscillator board of the WA7MLH 
receiver. The BFO is mounted on the right- 
hand wall of the box. 


The receiver is constructed on three 
circuit boards. These may be seen in the 
photographs. The LO is built on a board 
that is mounted close to the tuning 
capacitor. The slug-tuned inductor is 
mounted on a scrap of pc board that is 
soldered to the main board. The BFO is 
on a second board which is located on 
one of the side walls of the cabinet. The 
remainder of the receiver is on a larger 
board that is affixed to the rear wall of 
the receiver. 

All of the pc boards are double- 
sided, with one side serving as a ground 
plane. Coaxial cable (RG-174) is used 
for connections between boards and to 
the panel-mounted components. 

An aluminum plate is mounted to 
the bottom of the tuning capacitor. 
While this plate could serve as a chassis 
for some of the boards, its main func- 
tion is to isolate the receiver from 
additional circuitry. 

Considering its simplicity, this re- 
ceiver performs very well. A signal of 
0.1 yV from a well-shielded signal gen- 
erator was copied easily, indicating 
more than ample sensitivity. The selec- 
tivity of the two-pole filter is quite 
respectable for cw operation, and tire 
stability is compatible with the narrow 
bandwidth. No problems with overload 
or IMD products have been observed. 

A Superhet for 80 and 20 Meters 

There are a number of frequency 
schemes that lend themselves to simple 
two-band receivers. The previous super- 
het for 80 and 40 was one example. The 
unit shown in Fig. 74 is another. Here a 
9-MHz i-f is combined with a 5- to 
5.5-MHz LO in a receiver covering the 
80- and 20-meter bands. Another that 
might be interesting would be an 80- 



Front panel of the 80- and 20-meter superhet- 
erodyne receiver. Dial calibration is for the 
20-meter band. 


and 15-meter design. A 12- to 13-MHz 
oscillator would provide full coverage of 
both bands with a 9-MHz i-f. 

The front end of the 80/20 receiver 
uses a 40673 MOSFET mixer with no rf 
amplifier. Separate preselector networks 
are used for each band. A single-pole 
double-throw toggle switch is used to 
change bands at the output of the 
preselectors. Separate coaxial con- 
nectors are used at the input of each 
preselector, as the unit is used occa- 
sionally for 80-meter cw work, but was 
intended primarily as a tunable 14-MHz 
i-f system for use with vhf converters. 



Front-end section of the 80- and 20-meter 
receiver. The circuit board is mounted on the 
VFO tuning capacitor. The dual-section 
variable capacitor tunes the 80-meter pre- 
selector. The small single section variable is 

used with the 20 -meter input circuit. 


Receiver Design Basics 103 






33 33 



Fig. 74 - Schematic diagram of the 20- and 80-meter superheterodyne receiver. Fixed-value capacitors are disk ceramic unless noted. 


Fixed-value resistors are 1 / 2-Watt composition. 

C4 - Two-section 140-pF variable. 

CIO 100-pF variable (one section of BC-455 
variable). 

LI - 25 turns No. 28 enam. on T37-6 toroid 
core, 1 .87 pH. 

— 3 turns No. 28 enam. wire over LI . 

L5 — 44 turns No. 26 enam. wire on T68- 
2 toroid core. 1 0.8 pH. 

— 2 turns No. 26 enam. wire over L3. 

— 5 turns No. 28 enam. on T30-2 toroid 
core. 


Polarized capacitors are electrolytic. Numbered capacitors not listed below are trimmers. 


L7, L9 - 25 turns No. 28 enam. wire on 
T37-6 toroid core. 1.87 pH. 

L8 - 6 turns No. 28 enam. wire over L7. 

LI 0 — 4 turns No. 28 enam. wire over L9. 

L1 1 - 40 turns No. 28 enam. wire on T37-6 
toroid core. 4.8 pH. 

LI 2 — 3.5-pH inductor on ceramic form 
(Miller 4505 coil with slug removed). 
Remove turns for desired tuning range. 
L13 — 30 turns No. 26 enam. wire on an 
Amidon T50-6 toroid core, 3.5 pH 


R1 — Small 50,000-ohm carbon control. 

R2 — 20,000-ohm audio-taper carbon control. 

SI, S2 - Spdt toggle. 

S3 — Single-pole, three-position miniature 
switch. 

T1 — 12 trifilar turns No. 28 enam. wire on 
Amidon FT-37-61 ferrite toroid core, 
p = 125. 

VR1 — 6.8-volt, 1-watt Zener diode. 

Y1 — 9-M Hz crystal. International Crystal 
Co., type GP. 


bro 

tun 

80 - 

dou 

filte 

wid 


of l 
a t- 


104 Chapter 5 









I-F AMPLIFIER -hzv 


Vi-M. 


PRODUCT 

DETECTOR 


27k > -T- 009 


+I2V JIN9I4 HN914 


<0kV*— 
i-f >cw 
S*in5 



3.3juF 

560 '5V 

A A _ >1* 


X X 


4=^1 


4700 <£ -t- ?2 >< b 


09 


EXTERNAL 
1 +I2V 
TO MUTE 


,4700 S2 


>470 BFO 


XX 


08 

2N3565 


07 / n 

MPFI02 -£C8 


I 2N3906 S 10 , 


]Y1 >10k 

9MH>> 


X lQpF i 
,5v ^2200 


R1 1N914A 


f 50k 

:.47 'level 


oV" 


EXCEPT AS INOICATED, DECIMAL 
VALUES OF CAPACITANCE ARE 
IN MICROFARADS | jiF I ; OTHERS 
ARE IN PICOFARADS I pF OR ppFf 
RESISTANCES ARE IN OHMS; 
k *1000 . M* I 000 000 


AF AMPLIFIER 


1 N914 A ▼ XlN914A 


20k • 
AF GAIN 


Oil 

, TIP-298 


150S 

1W> J+ SOxF 
| T' 15V 


AF OUT 
(16 OHMS) 


Both preselector controls are 
Drought to the front panel. A single 
amed circuit is used at 14 MHz. For 
80-meter operation, an adjustable 
Jouble-tuned circuit was chosen. This 
Uler was designed for a 50-kHz band- 
mdtii and has a Butterworth response. 

The local oscillator is a FET version 
si the Colpitts circuit. It is followed by 
1 two-stage buffer amplifier using bi- 


polar transistors. A surplus tuning ca- 
pacitor from a BC455 is used for tuning. 
Only one section is employed. 

A single-sideband type of crystal 
filter is used as the basis of the i-f strip. 
Tlus is followed by an MC1350P IC i-f 
amplifier which supplies approximately 
45 dB of gain, and over 65 dB of gain 
variation. The filter and the IC amplifier 
are mounted on a small double-sided pc 


board which is buried in the chassis. 

The product detector uses two 
diodes. In spite of its simplicity, it 
performs well. The BFO employs a 
single transistor, and supplies +13 dBm 
ot injection to the detector. A single 
crystal was used, limiting reception to 
upper sideband or cw. Tire builder 
mi glit consider crystal switching if he 
wishes to copy lower sideband (pre- 

Receiver Design Basics 105 



Imerior of the 80- and 20-meier receiver. The 
5-MHz VFO is visible below the surplus tuning 
capacitor. The dual-section variable capacitor 
is part of the 80-meter preselector. At the 
lower left is a low-noise rf amplifier which is 
used in conjunction with some vhf converters. 


dominant for the 75-meter band) and 
for the output of OSCAR 7 (Mode B) 
on 2 meters. 

The major audio gain is provided by 
a pair of 2N3565s. The output of this 
amplifier is fed to the audio-gain con- 
trol. The audio-output amplifier uses a 
Darlington emitter follower to drive 
low-impedance stereo headphones. This 
provides excellent audio quality for 
reception of vhf phone signals. 

The high end of the audio-gain con- 
trol is sampled in order to drive the age 
detector. The age amplifier consists of a 
JFET and a pnp transistor in a low-gain 
feedback pair. The control is adjusted 
for an output of 5 volts with no signal 
present at the age detector. While the 
audio-derived age system suffers from 
the problems typical of such circuits, it 
is adequate for most ssb work. If the 
receiver is used for other than casual cw 
work, the builder might consider an i-f 
derived age detector. 

A single-pole, double-throw, center- 
off type of toggle switch is mounted on 
the front panel. In the center position, 
the receiver functions normally. When 
thrown in one position, a pair of back- 
to-back silicon diodes is inserted as an 
audio limiter. This helps considerably in 
attenuating the automotive ignition 
noise encountered on six meters. In the 
other position, the switch shorts the 
audio output for muting purposes. 

The photographs show a top view of 
the chassis. The LO board is mounted in 
back of the tuning capacitor while the 
front-end mixer is mounted on a vertical 
board which is bolted to the side of the 

106 Chapter 5 


tuning capacitor. The extra board con- 
tains a low-noise preamplifier for 14 
MHz. This is used in conjunction with a 
diode-ring mixer for vhf reception. With 
the “preamp,” the noise figure at 14 
MHz is on the order of 2 dB. For most 
20-meter operation this preamp is not 
necessary, since the mixer input pro- 
vides a system noise figure of about 10 
dB, which is adequate. 

A Unitized Receiver 
for 40 and 20 Meters 

Compactness is the key word in this 
superheterodyne design (Fig. 75). Cov- 
erage of 7000 to 7175 and 14,000 to 
14,175 kHz is available with this 
mini-receiver which operates from 12 or 
13 volts dc. Maximum current drain is 
120 mA, and idling current is on the 
order of 50 mA. The dimensions (HWD) 
are 2-5/8 X 4-3/4 X 5 inches. A minia- 
ture speaker is built in, and a speaker- 
disabling jack permits the use of head- 
phones. A minimum number of panel 
controls are used (tuning, band switch, 
and i-f gain) to make operation afield or 
at home as simple as possible. 

The basic receiver is a 40-meter 
superheterodyne. There is no age or af 
gain control. A simple single-crystal i-f 
filter is used to minimize cost and 
circuit complexity. The i-f bandpass is 
adequate for most cw work and is wide 
enough for ssb reception. 

Wide dynamic range was not the goal 
in this design. Rather, a sensitive and 
stable portable unit was desired, which 
led to some minor trading off in the 
performance features. However, for all 
but the most stringent applications, this 
unit is excellent. 

Coverage of the 20-meter cw band is 
effected by means of a simple two- 
transistor “down converter” which is 
mounted inside the main cabinet. 
Tuning on 20 meters is the reverse of 
that on 40 meters, owing to the crystal 
frequency used in the converter. If cw 
and ssb coverage is desired, the VFO 
tuning range will need to be extended. 
Furthermore, two BFO crystals will be 
necessary, plus a switch, to permit 
selecting upper or lower sideband. A 
0.1 -/aV signal is plainly audible on both 
bands. Since that level of sensitivity is 
greater than necessary for most work, 
an rf attenuator can be used between 
the antenna and receiver input to mini- 
mize mixer overloading. A simple 
brute-force attenuator will suffice a 
500-ohm carbon control between the 
mixer input link and ground, with the 
antenna connected to the control arm. 

Circuit Details 

T1 is designed to match a 50-ohm 
antenna to the 2000-ohm base-to-base 
impedance of the CA3028A balanced- 
mixer IC (Fig. 75). The transformer is 
broadband in nature (300 kHz at the 



Outside view of the unitized 40- and 20-meter 
receiver, dwarfed by a vacuum tube. The pc 
board in the foreground contains the 20 -meter 
converter. The cabinet is homemade and 
consists of two U-shaped pieces of aluminum 
stock. The front end rear panels are fashioned 
from double-sided pc board. Dimensions are, 
in inches. 2-5/8 X 4-3/4 X 5. Dymo-tape 
labels identify the controls. (From QST for 
September, 1976.) 


3-dB points) and has a loaded Q of 23. 
This eliminates the need for a front- 
panel peaking control - a cost-cutting 
aid to simplicity. 

The output tuned circuit, LI, is a 
bifilar-wound toroid which is tuned 
approximately to resonance by means 
of a mica trimmer, C2. The actual 
setting of C2 will depend upon the 
degree of i-f selectivity desired, and 
typically the point of resonance will not 
be exactly at 3300.5, the i-f center 
frequency. 

A single crystal filter with a phasing 
capacitor, C3, is used. This approach 
provides reasonably good single-signal 
reception (at least 30-dB rejection of 
the unwanted response) and assures 
much better performance than is pos- 
sible with the simpler direct-conversion 
receivers in vogue today. The latter have 
equal signal response each side of zero 
beat, which often complicates the QRM 
problem. 

A single i-f amplifier, U2, is used to 
provide up to 40 dB of gain. R1 serves 
as a manual i-f gain control, and will 
completely cut off the signal output 
when set for minimum i-f gain. T2 is 
designed to transform the 8000-ohni 
collector-to-collector impedance of U2 
to 500 ohms, and has a bandwidth of 
100 kHz. The loaded Q is 33. 

A two-diode product detector con- 
verts the i-f energy to audio. BFO 
injection voltage is obtained by means 
of a crystal-controlled oscillator, Q2. 
RFC2 and the 1-yF bypass capacitor 
filter the rf, keeping it out of the audio 
line to U3. 



lrhnni a rf^ h ^ a |! l o C H dia9ra 1 rn ° f ^ 4< ^. m 1 ! B . te . r rece,ue '- F'*ed-value capacitors are chip or disk ceramic unless noted otherwise. Capacitors 
Cl C2 C4 - n0 k to d 600^.F C mrM Irlm 3 ^' ' ndlCa,es Sllver m,ca ' and P is ,or Polystyrene. Fixed-value resistors are 1 /4- or 1 /2-W composition. 
C.,C2,C4 1 70 to 600-pF mica trimmer nommaLJ. W. Miller 42A105CBI or equiv. T-50-2 core. Turns rat,o - 6:1 , G, of 

. , '?*•. . „ 23. BWL = 0.3 MHz, L = 1 M H. 

L3 -Toroidal inductor, 17 uH. 19 turns No. 


Cl, C2, C4 — 1 70 to 600-pF mica trimmer 
(Arco 4213). 

C3 - 10-pF subminiature trimmer. Ceramic 
or pc-mount air variable suitable. 

C5 - Miniature air variable, 30-pF maximum 
(Millen 25030E or similar). 

CR1-CR3, incl. — High speed silicon switching 
diode 

J1, J3 - Single-hole-mount phono jack. 

J2 - Closed-circuit phone jack. 

U — Toroidal bifilar-wound inductor, O,. = 
100 at 3.3 MHz, Q, = 33, BWL = 0.1 
MHz, L = 5.8 pH. 8 turns No. 28 enam., 
bifilar wound on Amidon FT37-61 
ferrite core. Note polarity marks. 

L2 - Slug-tuned inductor (see text), 1 1 pH 


26 enam. wire on Amidon FT50-61 
ferrite core. 

R1 — 10,000-ohm miniature composition 
control, linear taper. 

RFC1, RFC2 — Miniature 1-mH choke 
(Millen J302-1000 or equiv.). 

RFC3, RFC4 - Miniature 330-pH rf choke 
(Millen J302-330 or equiv.). 

SI — Miniature dpdt toggle. 

T1 — Toroidal transformer. Primary has 2 
turns No. 24 enam. wire. Secondary has 
14 turns No. 24 enam. wire on Amidon 


T2 — Toroidal transformer. Primary has 9 
turns No. 26 enam. wire on Amidon 
FT37-61 core. O l - 33, BWL = 0.1 MHz 
L = 5.8 pH. turns ratio = 3.8:1 . Secondary 
has 3 turns No. 26 enam. wire. Primary 
winding has center tap. 

U1 — RCA 1C. Bend pins to fit 8-pin dual- 
inline 1C socket. 

U2. U3 — Motorola 1C. 

VR1 — Three-terminal 8-volt regulator 1C 
(National Semiconductor). 

Y1 , Y2 - Surplus crystal in HC-6/U case or 
International Crystal Co. type GP with 
32-pF load capacitance. 


Audio-output IC U3 contains a pre- 
amplifier and power-output system. It 
will deliver approximately 300 mW of af 
energy into an 8-ohni load. RFC5 is 
used to prevent rf oscillations from 
occurring and being radiated to the 
front end and i-f system of the receiver. 
The 0.1 -pF bypass at RFC5 also helps 
prevent oscillations. 

A three-terminal voltage regulator, 
VR1 , supplies the required operating 
voltage to U3. It also provides regulated 
voltage for the VFO and buffer stages of 
ihe local oscillator (Q2 and Q3). The 


latter consists of a stable series-tuned 
Clapp VFO and an emitter-follower 
buffer stage. A single-section pi network 
is placed between the emitter of Q3 and 
the injection terminal of Ul. It has a 
loaded Q of 1, and serves as a filter for 
the VFO output energy. It is designed 
for a bilateral impedance of approxi- 
mately 500 ohms. The recommended 
injection-voltage level for a CA3028A 
mixer is 1 .5 rms. Good performance will 
result with as little as 0.5-volt rms. A 
1-volt level is available with the circuit 
shown in Fig. 75. 


A red LED is used at DS1 as an 
on-olf indicator. Since it serves mainly 
as "window dressing," it need not be 
included in the circuit. 

Construction Notes 

The front panel, rear panel, side 
brackets, and chassis are made from 
double-sided circuit-board material. The 
chassis is an etched circuit board, the 
pattern for which is given in Fig. 77. 
There is no reason why the top and 
bottom covers for the receiver can not 
be made of the same material by sol- 

Receiver Design Basics 107 







Interior of the unitized receiver. The local oscillator is seen in its compartment at the center. 

A press-fit U-shaped cover is placed over the VFO box when the receiver is operating. The 
receiver front end is at the lower right. At the upper left is a miniature speaker, the rim of 
which is tack soldered to the box wall at four points. The 20-meter converter board mounts on 
the rear wall of the box (upper left). 



Fig. 76 - Schematic diagram of the 20-meter 
ceramic unless noted otherwise. Resistors are 
C6, C7 — 40-pF subminiature ceramic 
trimmer. 

J4 — Single-hole-mount phono jack on 
rear panel of main receiver. 

L4 — Toroidal inductor, 12 turns No. 26 
enam. wire on Amidon FT37-61 core 
Q l = 14, BWL = 0.5 MHz, L = 8 pH. 

L5 — Toroidal inductor. 24 turns No. 26 
enam. wire on Amidon T-50-6 core. 

Q u » 200 at 7.9 MHz. L - 2.4 pH. 

Q4 — RCA transistor. 


converter. Fixed-value capacitors are disk 
1/4- or 1/2-W composition. 

Q5 - Motorola transistor, MPF102, 2N4416 
or HEP802. 

T3 — Toroidal transformer, 10:1 turns ratio. 
Ql = 46. BWL - 0.3 MHz. L = 1 .85 pH. 
Pri. has 2 turns No. 26 enam. wire. Sec. 
contains 21 turns No. 26 enam. wire on 
Amidon T-50-6 core. 

Y3 - 21 .1 75-MHz fundamental crystal in 
HC-18/U case (International Crystal 
Co. type GP with 32-pF load capacitance). 


tiering six pieces of pc board together to f 
form two U-shaped covers. 

The local oscillator is housed in a 
compartment made from pc -board 
sections. It measures (HWD) 1-3/8 X 
1 -S/8 X 2-3/4 inches. A 1 /4-inch high 
pc-board fence of the same width and 
depth is soldered to the bottom side of 
the pc board (opposite the VFO top- 
chassis compartment) to discourage rf 
energy from entering or leaving the local 
oscillator section of the receiver (rf 
doesn’t like to climb over right-angle 
barriers). Employment of the top and 
bottom shields stiffens the main pc 
board, and that helps prevent mechani- 
cal instability of the oscillator which 
can result from stress on the main 
assembly. 

Silver plating has been applied to the 
main pc board, and to the front and rear 
panels. This was done to enhance the 
appearance and discourage tarnishing of 
die copper. It is not a necessary step in 
building the receiver. The front panel 
has been sprayed with green paint, then 
baked for 30 minutes by means of a 
heat lamp. A coarse grade of sandpaper 
was used to abrade the front panel 
before applicadon of the paint. The 
technique will prevent the paint from 
coming off easily when the panel is 
bumped or scratched. Green Dymo tape 
labels are used to identify the panel 
controls. 

There is ample room inside the 
cabinet, along the rear inner panel sur- 
face, to install the 20-meter crystal- 
controlled converter. A switch, SI, is 
located on the front panel to accom- 
modate a 20-meter converter, the circuit 
for which is given in Fig. 76. 

All of the toroidal inductors are 
coated several times with Q dope after 
they are installed in the circuit. The 
VFO coil is treated in a like manner. 
The polystyrene VFO capacitors should 
be cemented to the pc board after the 
circuit is tested and approved. This will 
help prevent mechanical instability. 
Hobby cement or epoxy glue is okay for 
the job. Use only a drop or two of 
cement at each capacitor - just enough 
to affix it to the pc board. 

Alignment and Operation 

The VFO should be aligned first. This 
can be done by attaching a frequency 
counter to pin 2 of U1 . Coverage should 
be from 3699.5 to 3874.5 kHz for recep- 
tion from 7.0 to 7.175 MHz. Actual 
coverage may be more or less than the 
spread indicated, depending on the 
absolute balues of the VFO capacitors 
and stray circuit inductance and capac- 
itance. Greater coverage can be had by 
using a larger capacitance value at C5, 
the main tuning control. Those in- 
terested in phone-band coverage (only) 
can align the VFO accordingly and 
change Y2 to 3302.3 kHz. 


108 Chapter 5 




Final tweaking is effected by at- 
taching an antenna and peaking Cl , C2 
and C4 for maximum signal response at 
7087 kHz To obtain the selectivity 
characteristics desired (within the capa- 
bility of the circuit), adjust C2 and C3 
experimentally. C2 wiil provide the 
major effect. C3 should be set for 


minimum response on the unwanted 
side of zero beat. A fairly strong signal 
will be needed to hear the unwanted 
response. 

For reception of lower sideband it 
will be necessary to use a different BFO 
frequency - 3298.7 kHz. The crystal 
indicated in Fig. 75 was used because 



ground-plane side ot board opposite LI, C3 and Y1 (1 X 1-1/4 inch area). Removal of foil 
will prevent unwanted capacitive effects. The 100-kr gate 2 resistor is on etched foil of 
board, gate 2 to source. Ground-plane side of board should be electrically common to 
ground foils on opposite side of board at several points. 


only cw reception was intended. Those 
wishing to shift the BFO frequency a 
few hundred Hz can place a trimmer in 
series with Y2 rather than use the 
100-pF capacitor shown. 

Because there is no age in this 
receiver, the i-f gain should be set low, 
for comfortable listening. Too much 
gain will cause the audio circuit to be 
overdriven. and distortion will result. To 
prevent ear-splitting signal levels one can 
install a pair of IN34A diodes (back to 
back) across the output jack, J2. 

Bits and Pieces 

The photograph shows some 
fancy-looking components on the 
circuit board. Tantalum capacitors are 
seen where electrolytics are indicated 
on the diagram. Either type will work 
nicely. Tantalums were found at a 
flea market for 10 cents each, so 
they were used. Similarly, the 0.1 -/jF 
capacitors used are the high-class kind 
(Aerovox CK05BX) which sell for 
roughly 70 cents each. At the Ilea 
market they sold at $1 for 44 pieces! 
Mylar or disk ceramic 0.1 -,uF units 
will be fine as substitutes. 

The polystyrene capacitors were ob- 
tained from Radio Shack in an assort- 
ment pack. New units are made by 
Centralab, and they sell for less than 20 
cents each in single lots. Since they are 
more stable than silver micas, they are 
recommended for the VFO circuit. All 
of the toroid cores were purchased by 
mail from Ajnidon Associates. 

A J. W. Miller 42 -series coil is used in 
the VFO, but any slug-tuned ceramic 
form can be used if it has good high- 
frequency core material. The unloaded 
Q of the inductor should be at least 150 
at 3.5 MHz. L2 in this design has a 
3/8-inch diameter body. The winding 
area is 5/8 inch long. 

The metal cases of both crystals 
should be connected to ground by means 
of short lengths of wire. This will prevent 
unwanted radiation from the BFO crys- 
tal. and will help keep the filter crystal 
from picking up stray energy. A metal 
cover should be placed on the VFO 
compartment for reasons of isolation. 

James Millcn encapsulated rf chokes 
are used in the receiver. Any sub- 
miniature choke of the approximate 
inductance indicated will be suitable, 
and it need not be encapsulated. The 
VFO tuning capacitor is also a Millen 
part. Ample room exists between the 
VFO box and the front panel to allow 
making the box longer. That will permit 
use of a larger variable capacitor. A 
double-bearing capacitor is recom- 
mended for best mechanical stability of 
the VFO. 

The i-f system and BFO can be 
tailored to frequencies other than those 
indicated. If crystals of other fre- 
quencies in the 2- to 3-MHz range are 


Receiver Design Basics 109 



chosen, the VFO, mixer, and i-f ampli- 
fier tuned circuits will have to be altered 
accordingly. 

No hum or distortion is heard in the 
output of the receiver at normal lis- 
tening levels. VFO drift is 45 Hz from a 
cold start to stabilization, and strong 
signals do not pull the oscillator. 

Extremely strong local signals (1000 
/aV or greater) will cause desensitization 
of the receiver when they appear off 
frequency from where the operator is 
listening. Under ordinary conditions this 
will not be a problem. At some sacrifice 


in noise figure and sensitivity, those 
living in areas where other amateurs are 
nearby can modify T1 to aid the situa- 
tion. Cl remains across all of the T1 
secondary, and a 2200-ohm resistor is 
paralleled with Cl. Pins 1 and 5 of U1 
should be connected two turns each side 
of the center tap of the secondary. This 
will require cutting the pc -board ele- 
ments to divorce pins 1 and 5 from Cl. 
This design tradeoff is quite acceptable 
at 40 meters, as the atmospheric noise 
level will mask the reduction in receiver 
noise performance. With the circuit 


change there was no desensing evident 
below approximately 8000 /aV. 

Age could be used in this receiver by 
applying an audio-derived type. If the 
feature were adopted, age voltage would 
be applied to pin 5 of U2 and the 
manual gain control would be elimi- 
nated. In such a case it would be 
necessary to add an af-gain control 
between the product detector and U3. 
It should be remembered that minimum 
gain results when 13 volts are applied to 
pin 5 of U2. The lower the voltage at 
that point, the greater the gain. 


110 Chapter 5 


Chapter 6 


Advanced Receiver Concepts 


Q 

Uome fundamentals ol receiver de- 
sign were presented in chapter 5. How- 
ever. there was minimal discussion of 
receiver front-end design. That infor- 
mation forms the basis for most of this 
diapter. 

Conditions in the amateur bands are 
much different than they were even ten 
years ago. The spectrum is crowded, 
with demands for additional space 
arising daily. Furthermore, the power 
levels are increasing. In the past it was 
only the occasional amateur that ran the 
full legal power limit. Today kilowatt 
amplifiers are common. 

These conditions call for better re- 
ceivers than those used in the past. Not 
only must selectivity, sensitivity and 
stability be maintained, but the receiver 
must meet these specifications while 
operating in the presence of numerous 
strong signals. We will present informa- 
tion tn this chapter that will help the 
amateur experimenter to meet these 
goals. 

The critical portion of the receiver is 
the front end, that part which precedes 
the main selectivity-determining ele- 
ments. Distortion effects in the front 
end will lead to blocking, intermodula- 
tion products and cross modulation. 
Careful design is necessary if these 
phenomena are to be minimized. 

Dynamic Range 

In the previous chapter, some of the 
basic specifications of receivers, in- 
cluding the idea of noise figure, were 
outlined. Implicit in the noise-figure 
concept was the fact that the minimum 
discernable signal (MDS) of a receiver is 
dependent not only upon the amount of 
noise generated by the transistors in the 
receiver, but upon the bandwidth of the 
system. 

While sensitivity is of major signifi- 
cance to the amateur with an interest in 
DXing. a receiver must be able to 


survive in the presence of strong signals. 
This has a twofold meaning. First, the 
gain-control mechanisms in the receiver, 
manual or automatic, must have a range 
that will permit signals with wide 
strength variations to be received. How- 
ever, this can be realized easily — in the 
extreme case, attenuators in the antenna 
line can be used to decrease the signal 
level to a point where intelligence can 
be recovered. 

The second, and more subtle figure 
for dynamic range, is a number which 
provides a measure of the range of 
signals which may be present at the 
antenna terminals of a receiver while no 
undcsired responses are created in the 
output. The various ways that such a 
range can be defined, and the way it is 
measured, are described in this section. 
Also, we will show how the concepts 
surrounding these measurements may be 
utilized in the design of a receiver. 

Consider a simple amplifier in the rf 
or i-f portion of a receiver. For our 
example, we will assume that the ampli- 
fier uses a bipolar transistor and is 
biased for a collector current of 10 mA. 
The concepts are applicable to any 
amplifier, mixer or complete receiver. 

First, we will consider the mea- 
surement of the noise figure of the 
amplifier. By definition, the noise factor 
of the amplifier is the input signal-to- 
noise ratio divided by the output signal- 
to-noise ratio 

NF- /'"f --- =1 =%' (FO-D 

^oull^our •» our'in 

The terms in the equations are noise or 
signal powers, and the noise factor is an 
algebraic ratio. If we express that ratio 
in dB, as is often done with other power 
ratios (e.g., gains), the result is the noise 
figure. 

As presented, the noise figure is a 
nebulous number, for the input (and 


hence, the output) noise power is de- 
pendent upon what is hooked to the 
input of the amplifier. In order to 
attach some meaning which will make a 
noise figure number a standard measure 
of the “noisiness" of an amplifier or 
receiver, the input noise is assumed to 
be the noise power available from a 
resistor at a temperature of 290 degrees 
Kelvin. Using this value for T„, the 
noise power is given as P„ = kT„B, 
where T a = 290 degrees Kelvin, B is 
bandwidth in Hz. and k is Boltzman’s 
constant, 1.38 X 10' 23 watts/degree. It 
is convenient to use lo^rithmic units 
and to note that in a bandwidth of 1 
\\z,P„ = —174 dBm. 

Consider a receiver with a bandwidth 
of 500 Hz. Tlw bandwidth is greater 
than one Hz by a factor of 500, or 27 
dB. Hence, in a 500-Hz bandwidth, the 
power available from this resistor would 
be 174 dBm + 27 dB = 147 dBm. If 

the noise output from this receiver witli 
the input terminated in a 50-ohm resis- 
tor corresponds to that output which 
would result from a signal of 140 
dBm, the noise figure of the receiver is 
then the difference, or 7 dB. The MDS. 
or noise floor of the receiver is 140 
dBm. 

One might ask why noise figure is 
even specified. The same essential infor- 
mation is contained in a specification of 
the MDS of a receiver. However, such is 
not the case for an amplifier. Here, the 
MDS is not specified - it will depend 
not only upon the noise contribution of 
the amplifier, but on the bandwidth of 
the system using that amplifier. Noise 
figure is independent of bandwidth. 

A further asset of noise figure is that 
it is, at least in principle, measured 
easily. This is a direct result of the 
bandwidth invariance. The measurement 
is performed by attaching a source to 
the input of a receiver (or amplifier) 
that has a noise output which is known 


Advanced Receiver Concept in 



Fig. 1 — The principle of noise temperature. 


to be some well-defined factor greater 
than tljat of a room -tempera lure resis- 
tor. As long as the noise from this 
source is distributed evenly (white 
noise) over the frequencies of interest, 
the device being measured will respond 
to this known input with exactly the 
same filtering bandwidth that is applied 
to the internally generated noise. By 
measuring the increase in output noise, 
the noise figure is easily calculated. 
Knowledge of the system bandwidth is 
not required in the calculation. 

A related concept which also de- 
scribes the noisiness of an amplifier or 
receiver is that of noise temperature. 
This concept is outlined in Fig. 1 , where 
the device being evaluated is modeled 
by a noiseless amplifier preceded by an 
“ideal adder" and a noise generator. The 
excess-noise generator represents the 
noise that is contributed by the ampli- 
fier. Effective noise temperature is re- 
lated to noise factor by the equation 
F - 1 + T ef/ -IT 0 , where T e j f is the 
effective noise temperature of the am- 
plifier and T a is the reference tempera- 
ture (usually 290 degrees Kelvin). This 
equation is derived easily if we recall 
that noise factor can be expressed as the 
ratio of noise gain to signal gain. If the 
available gain of the amplifier is G, the 
noise output will be 

N out = G(kT a B + kT e/f B). (Eq. 2) 


The input noise power is just kT 0 B. 
Noise factor is then 


G n - 1 ,. D (To + T eff ) 
G s ~G kH 


kBT 0 



(Eq. 3) 


As an example, assume that the 
effective noise temperature of an ampli- 
fier is 400° Kelvin. The noise factor is F 
= 1 + 400 -r 290 = 2.38. The noise figure 
is 3.76 dB . 

The advantage of the noise- 
temperature concept over that of noise 
figure is that it is an absolute number. It 
is not dependent upon the more or less 
arbitrary choice of a reference tempera- 
ture. It also has the advantage that it is 
in some cases, a more meaningful in- 


112 Chapter 6 


dicator of the ability of a system to 
detect very weak signals. This requires 
some elaboration. 

Assume that a receiver with a noise 
figure of 3 dB is made more sensitive by 
adding a preamplifier which provides a 
net system-noise figure of 0.5 dB. One 
might assume that because tire noise 
figure of the system has decreased by 
2.5 dB, we will be able to hear signals 
which are 2.5 dB lower. However, this is 
generally not the case. It would be true 
only for the situation where the input 
noise to the system was originating from 
a 290° Kelvin source. If the noise was 
originating from atmospheric distur- 
bances (causing noise in the hf spec- 
trum), the increase in output signal -to- 
noise ratio would be virtually imper- 
ceptible. On the other hand, if the noise 
was from a large parabolic antenna 
pointed toward one of the quieter parts 
of outer space, the input noise would be 
nearly zero. In this case, the 2.5-dB 
improvement in noise Figure could lead 
to an approximate 9-dB improvement in 
receiver sensitivity. This conclusion re- 
sults from Eq. 3, which shows that a 
drop in noise figure from 3 to 0.5 dB 


corresponds to decreasing the effective 
noise temperature from 290° to 35° 
Kelvin. 

Consider now the case where two 
relatively strong signals are placed simul- 
taneously at the input to the 20-dB 
amplifier mentioned earlier. Assume 
that two input signals of -50 dBm are 
placed at the input of the amplifier at 
frequencies /, and f 2 . Analysis of the 
amplifier using mathematics outlined in 
the appendix will show that distortion 
in tire amplifier will give rise to outputs 
not only at die desired input fre- 
quencies of and fi , but at (2/, / 2 ) 

and (2/j - /,). For example, if die 
input frequencies were 14,040 and 
14,050 kHz, the distortion products 
would appear at 14,030 and 14,060 
kHz. In die amplifier the desired out- 
puts would be 20 dB above die 
-50-dBm input signals, or —30 dBm. 
and the 3rd-order distortion products 
would be at —130 dBm. In this case the 
distortion will be 100 dB down from 
the desired outputs. 

The interesting and significant char- 
acteristic of Class A linear amplifiers is 
that while the desired outputs will vary 




Fig. 2 - Plot example showing signal power versus distortion products as a function of input 
power of two identical input signals. 






linearly with changes in the input sig- 
nals. the dominant distortion products 
will vary as (lie cube of the input 
powers. Hence, if we increase the signals 
driving the input to -40 dBm. the 
output power of the desired signals will 
be 20 dBm for each of the desired 
input tones. However, while the level of 
the desired frequencies increased by 10 
dB, the output pcwer of the distortion 
products will have increased by 30 dB 
to - 100 dBm. The distortion products 
are now only 80 dB below the desired 
results. 

Shown in Fig. 2 is a plot for our 
hypothetical amplifier, showing the 
power of the desired output signals and 
the output power of the distortion 
products as a function of the level of 
the input power of each of the two 
identical input signals. Eventually, the 
level of the input signals will be large 
enough so that the desired outputs cease 
to follow the input power linearly. This 
effect is called gain compression, and is 
the phenomenon in a receiver which 
ultimately leads to “blocking." It is not 
viable to plot the data for the amplifier 
much beyond this compression point. 

The linear portions of the curves 
may be extended, or extrapolated to 
higher powers even though the amplifier 
is not capable of operating at these 
levels. If this is done, as is shown in a 
dotted line in the figure, eventually the 
two curves will cross each other. That is 
at some usually unattainable output 
power, the level of the distortion pro- 
ducts equals that of the desired outputs. 
This point is commonly referred to as 
the amplifier intercept. More specifi- 
cally, the output power where the 
curves intersect is called the output 
intercept of the amplifier. Similarly, the 
input power corresponding to the point 
of intersection is called the input inter- 
cept. 

It is important to distinguish be- 
tween the input and the output inter- 
cepts when specifying a given device. In 
any useful amplifier (one with power 
gain) the output intercept is always 
greater than the input intercept by an 
amount corresponding to the gain of the 
amplifier. But with lossy circuits (such 
as a diode mixer) the input intercept 
will exceed the output intercept. In 
professional literature the number 
usually given is the output intercept. 
However, the input intercept is an 
equally important number when dis- 
cussing receivers. 

The value of knowing the intercept 
of an amplifier is that it is a general 
measure of the distortion properties. It 
can be used to describe the distortion 
for all operating levels. In the case just 
depicted the output intercept is +20 
dBm. Hence, if the amplifier is operated 
with an output which is X dB below the 
intercept, the distortion will be 3X 


below the intercept. For example, if the 
amplifier is operated with outputs of 0 
dBm, which is 20 dB below the inter- 
cept. the distortion products will be 
three times the 20-dB difference, or 60 
dB below the intercept at 40 dBm. In 
our example amplifier, the input inter- 
cept is 0 dBm. The same relationships 
apply using this figure of merit. 

It is generally not viable to specify 
the output intercept of a receiver, for 
this is a function of the gain setting of 
the unit. However, such is not always 
the case, with an input intercept. This 
number may be specified and is an 
extremely useful general parameter. 
Suppose, for example, that the input 
intercept of a receiver is 0 dBm. (This 
number is not purely arbitrary, but is 
representative of a well-designed com- 
munications receiver.) This means that 
if two signals are placed at the antenna 
terminals with levels of -40 dBm, the 
response when the receiver is tuned to 
the frequencies of the distortion prod- 
ucts (2/j f 2 or If 2 - f\ ) will be three 
times 40 dB below the input intercept, 
or the same as an input signal of — 120 
dBm. 

As is usually the case with receivers, 
the analysis of performance is compli- 
cated by noise. If the two inputs just 
mentioned were dropped to 60 dBm 
(which is 60 dB below the input inter- 
cept) the response at the distortion- 
product frequencies would be 180 dB 
below the input intercept or at -180 
dBm equivalent input signal. If this 
receiver had an exceptionally low noise 
figure and a bandwidth of a fraction of 
one Hz, this level of signal could be 
detected. However, this is not usually 
the case with communications receivers. 
If the receiver had a more typical MDS 
or noise tloor of -140 dBm, the dis- 
tortion products would not be de- 
tectable. This brings us to the concept 
of dynamic range. 

The two-tone dynamic range of a 
receiver is defined as the ratio of the 
noise floor (MDS) of the receiver to the 
level of one of two identical input 
signals which will cause distortion prod- 
ucts at the noise floor level. This con- 
cept is illustrated by considering a 
measurement on the receiver described 
in the foregoing discussion. 

First, the instrumentation is 
gathered and interfaced with the re- 
ceiver. This includes a pair of signal 
generators with means for combining 
their outputs while minimizing inter- 
action between them, and an ac volt- 
meter to monitor the audio output 
signal. 

The initial measurement uses only a 
single signal source. The generator is 
adjusted so that the output of the 
receiver is 3 dB above the level present 
when the generator is turned off. The 
power output (available output power 


in dBm) of the generator is then the 
MDS of the receiver. 

After measuring the receiver MDS, 
the two generators are set up for IMD 
measurements. The two generators are 
added in a 6-dB hybrid combiner. The 
output is applied to a step attenuator 
and then to the receiver. The attenua- 
tor is adjusted until the responses at the 
tliird order IMD frequencies are the 
same as that produced by the MDS. The 
DR in dB is then the dB difference 
between the power in each tone avail- 
able to the receiver input and the MDS. 

The two-tone dynamic range of a 
receiver is related to the input intercept 
of the receiver by the relationship 

Dynamic range (in JB)= 2/3(7', MDS) 

(Eq.4) 


where the input intercept, P h and MDS 
are in dBm. 

At the time that the receiver is being 
evaluated for intermodulation dis- 
tortion, blocking measurements are also 
performed easily. This is done by setting 
one of the generators to provide a 
medium-strength signal in the receiver . 
With the receiver tuned to this output, 
the other generator is increased in out- 
put until the desired output is reduced 
by I dB. This onset of desensitization, 
when compared with the noise tloor of 
the receiver, might be referred to as a 
“single-tone dynamic range." 

The use of blocking, and more 
specifically, intermodulation distortion 
as the mechanisms to define the strong 
signal performance of a receiver, might 
appear esoteric and restrictive. However, 
such is not the case. The blocking 
measurement will tell the user how well 
his receiver will survive when subjected 
to a strong neighbor. The two-tone 
dynamic range will indicate the level of 
signals which the receiver will tolerate 
while producing essentially no undesired 
responses. 

The authors have evaluated a num- 
ber of commercially built receivers. The 
best unit studied at this writing had a 
two-tone dynamic range of 88 dB with a 
noise figure of 5 dB. The single tone 
dynamic range was only I 16.5 dB.This 
unit used tubes in the front end. An 
"average” performer yielded two-tone 
and single-tone dynamic ranges of 80 
and 109 dB, respectively. On the other 
hand both authors have constructed 
solid-state receivers with two-tone dy- 
namic ranges approaching 100 dB. 
single-tone ranges of over 120 dB and 
noise figures from 6 to 13 dB. While 
sophisticated instrumentation was used 
for evaluation, both units were built 
using only equipment available in many 
amateur shops. Both receivers are de- 
scribed in this book. 

Advanced Receiver Concept 113 


\/ 


It is interesting to consider the effect 
of cascading two or more amplifiers (or 
a receiver with a “preamp” or con- 
verter) with respect to the effect on 
noise figure and dynamic range. 
.Knowing these, we will be able to 
calculate the resulting dynamic range. 

Consider two cascaded amplifiers. If 
they have noise factors /•', and F 2 , and 
gains G| and G 2 (both are algebraic 
ratios, not dB relationships) the net 
noise factor of the combination will be 
given by 


Fne, =F i +(F 2 - I)/G, (Eq.5) 


For example, assume that each amplifier 
has a gain of 20 (13 dB), that the first 
one has a noise factor of 2 (3 dB) and 
the second has a noise factor of 5 (7 
dB). The net noise factor is F„ et = 2 + 
(5 I ) -r 20 = 2.2, which corresponds 
to a noise figure of 3.42 dB. The net 
gain is 400, or 26 dB. Note that the net 
noise figure is dominated by the first 
stage of the amplifier if the gain of the 
first stage is large in comparison to the 
noise figure of the second stage. But. 
excess gain in the first stage beyond this 
level does little to improve the net noise 
figure. 

Assume that the first stage has an 
output intercept of +15 dBm and that 
tire second stage is stronger, with an 
output intercept of +20 dBm. Since the 
gain of the second stage is 13 dB. die 
input intercept of the second stage will 
be +20 13 = +7 dBm. Noting that this 

input-intercept amount is less than the 
output intercept of the first stage (a 
margin of 8 dB). the IM response of the 
composite amplifier will probably be 

) dominated by the distortion in the 

second stage. We can estimate the out- 
put intercept of the combined amplifier 
to still be +20 dBm. Since the overall 
gain is 26 dB. the input intercept of the 
cascaded pair will be +20 - 26 = — 6 
dBm. 

It should be mentioned that the IM 
distortions from two cascaded stages 
will add in a simple manner, with the 
output stage usually being the dominant 
contributor. However, there are some 
situations where the IM from one stage 
will add in a phase-coherent way with a 
following stage: The overall result is IM 
which is much worse than anticipated. 
In rare examples the opposite effect will 
occur, yielding better distortion prop- 
erties than predicted. These cases do not 
lend themselves to easy analysis or 
duplication. 

Note that in the foregoing discussion 
nothing has been said about dynamic 
range. This is because the dynamic range 
is defined while using an input equiva- 


lent noise floor which is dependent 
upon noise figure and system band- 
width. A dynamic range can be specified 
only when a bandwidth is given simul- 
taneously. 

As an extension of the discussion, let 
us consider adding a preamplifier to a 
receiver which is lacking in noise figure. 
Assume that the receiver has an excep- 
tionally poor noise factor of 100 (20 
dB), and a dynamic range of 80 dB. The 
bandwidth of the receiver is 500 Hz. 
The minimum detectable signal, or noise 
floor of the receiver will be 


Noise floor = — 174 dBm + noise figure 
+bandwidlh factor 
= -174 dBm + 20 dB 
+27 dB 
= -127 dBm 


(Eq. 6) 


If this receiver was to be used in the 
10-meter band a much lower noise 
figure might be in order. Assume that a 
preamplifier with a 3-dB noise figure is 
added. Following the earlier argument 
about noise figure, a preamplifier gain 
of 20 dB. equal to the receiver basic- 
noise figure, is used. The net noise 
figure becomes 


F = 2 +-? 2 - 
100 


= 2.99 or 4.76 dB 


(Eq. 7) 


The noise floor decreases to 


Noise floor = -174 dBm + 4.76 + 27 

= 142.24 dBm (Eq. 8) 

The improvement in sensitivity is pro- 
found. 

Consider now the effect of the pre- 
amplifier on the dynamic range of the 
receiver. Using the formula relating 
dynamic range’ to noise floor and input 
intercept, we deduce that (he input 
intercept of the basic receiver is -7 
dBm. If the preamplifier is even rea- 
sonable (from a distortion point of 
view), the distortion properties of the 
overall system will be dominated by the 
receiver basic input intercept of 7 
dBm. The system input intercept will be 
—27 dBm. The overall system dynamic 
range is 

Dynamic range = 2/3 ( 27 + 142.24) 

= 76.8 dBm (Eq. 9) 


The dynamic range has been slightly 
degraded from the original dynamic 
range of 80 dB. which is an acceptable 
compromise. 

If, however, the gain in the preamp- 


lifier had been set at 30 dB instead of 
the 20-dB level chosen, the extra gain 
would drop the noise floor to 143.78 
or merely 1.5 dB more sensitive. How- 
ever. the input intercept would drop to 
37 dBm, resulting in a dynamic range 
of 71.2 dB. Such a compromise would 
not be acceptable except perhaps in 
very rare situations such as moonbounce 
work on 144 or 432 MHz where noise 
figure is all! The price to be paid is 
always a severe degradation in dynamic 
range. 

One final comment should be made 
about receiver dynamic range. The com- 
mon “cure" dial is suggested for a 
receiver plagued with problems of over- 
load and excessive intermodulation dis- 
tortion is the addition of an attenuator 
in front of the receiver. Often this is an 
excellent thing to do. The attenuator is 
adjusted until the antenna noise still 
determines the overall noise output but 
is not excessive. 

The addition of a 10-dB pad in front 
of a receiver has the effect of increasing 
die system noise figure and the input 
intercept by 10 dB. The difference 
between the two, and hence the system 
dynamic range, remains constant. A 
much better solution would be to by- 
pass the offending amplifier, allowing 
smaller signals to impinge .upon die 
mixer. While die noise figure will be 
compromised, the dynamic range will 
usually be improved. 

There is anodier technique that may 
be applied to regain some of the system 
dynamic range: the application of atten- 
uation between the preamplifier and die 
main receiver. Consider the previous 
case where a mediocre receiver was 
preceded by a 30-dB-gain preamplifier, 
causing a net dynamic range of only 
71.2 dB. If a 12-dB attenuator was 
inserted between the preamplifier and 
the receiver, the net system MDS would 
increase from 143.8 to -141.5 dBm. 
However, the dynamic range would in- 
crease to 77.6 dB. This technique could 
be of major significance when building a 
dual-conversion system with crystal- 
controlled converters ahead of a tunable 
i-f receiver. 

While it is dangerous to generalize.it 
is clear that the optimum dynamic-range 
systems will be those utilizing single 
conversion. However, wide dynamic 
range is certainly possible in multicon- 
version designs. Great care must be 
applied in tailoring the gain distribution 
properly, in order to optimize the trade- 
off between dynamic range and noise 
figure. Careful measurements, as well as 
detailed calculations during the design 
phase, are mandatory. 

In the following sections, the design 
of mixers, amplifiers and filters will be 
considered in more detail than pre- 
sented in chapter 5. The major differ- 
ence iti this approach will be our inclu- 




114 Chapter 6 



Fig. 3 - Representation of a receiver input 
circuit, coupled capacitively. 


sion of intercept data as well as noise 
performance of the various devices. 

Preselector Design 

The previous section outlined the 
concepts of dynamic range and de- 
scribed some of tire undesired effects 
that arise from excessively strong signals 
appearing at the input of a receiver. 
Much of the key to minimizing these 
effects lies in the design of the mixers 
and amplifiers that make up the front 
end of a receiver. As much as possible 
should be done to ensure that the 
front-end components are subjected to a 
minimum of strong signals. This is rea- 
lized with careful filtering at the anten- 
na terminal of a receiver. Such a filter is 
called a preselector. 

The subject of filter synthesis is a 
complicated one. Sophisticated mathe- 
matics are required, making a complete 
discussion impractical in this book. 
However, some of the basic ideas can be 
presented. An extensive catalog of com- 
puter-designed filters for the amateur 
bands is given in the appendix for use in 
specific projects. 

The Single Tuned Circuit 

With most receivers in use today, the 
preselector consists of nothing more 
than a single tuned circuit preceding the 
rf amplifier (if one is used) or the mixer. 



REAL L AND C L AND C 


R P _ 2irf 0 L 
q ~2itf 0 L R, 

where F a = \= 

2*yfLC 

f ig. 4 - Modeling of an ideal resonator with 
series or parallel resistance. 


While this may be adequate to provide 
marginally acceptable image rejection, it 
usually provides a minimum of protec- 
tion from out-of-band signals that might 
lead to 1MD products. We will investi- 
gate this type of preselector for two 
reasons. First, the inadequacy of such a 
circuit will be demonstrated. Of more 
significance, we will use the single-tuned 
circuit to demonstrate some fundamen- 
tals that are applicable to any preselec- 
tor. 

Consider a receiver with the first 
semiconductor device having an input 
impedance of 50 ohms. If a preselector 
is to be designed for this receiver, it 
must be a circuit that is terminated on 
both sides (input and output) with a 
50-ohm load. A typical circuit is shown 
in Fig. 3 where capacitive coupling is 
used at both terminals. 

The concept of Q was introduced in 
our discussion of tuned transmitter buf- 
fer amplifiers. Q is a number that gives 
us information about the losses in a 
resonator. (The term resonator will be 
used interchangeably with “tuned cir- 
cuit.” The concepts are applicable to 
microwave resonant circuits just as they 
are to low-frequency LC tuned circuits 
and even to nonelectrical oscillations.) 
While Q tells us the amount of energy 
that is lost during each cycle of oscilla- 
tion, we can model a real resonator by 
replacing it with an ideal lossless one 
with either a parallel or series resistance. 
This is shown in Fig. 4 along with the 
equations which define the resistances. 

If the resonator exists alone, at- 
tached to no external load, the Q is the 
unloaded value, designated Q u . The 
associated resistances model the inher- 
ent losses within the inductor and ca- 
pacitor. In the high-frequency region 
inductive losses are predominant in 
most cases. Hence, one will often see a 
Q u specification for a coil at a given 
frequency. 

If external resistances are attached 
to the resonator, the resulting Q is 
termed the loaded value and is repre- 
sented by Ql . The corresponding resis- 
tance is the equivalent of all of the 
loads, including that representing the 
inherent resonator losses. 

A term that is rarely used but can 
occasionally be useful in calculations is 
Q e , the external Q. This is merely the Q 
associated with the external resistances 
attached to the tuned circuit. 

Let us now return to the filter 
described in Fig. 3 and consider the 
effect of the finite unloaded Q of the 
resonator. This is done by substituting 
the model of Fig. 4 for the tuned 
circuit, new shown in Fig. 5. First, there 
will be loss associated with this filter. If 
the filter was removed completely, with 
a direct connection between the source 
and load resistors (which here are 
equal), the power that would be de- 


livered to R l would be the maximum 
available amount that the generator 
could deliver. Substitution of the filter 
places another resistive element into the 
circuit. This is the loss resistance, R u , 
associated with Q u of the resonator. 
Since a voltage will appear across the 
resistor, it must dissipate power. This 
will be subtracted from the maximum 
available power from the generator. 

The loaded Q of the resonator is 
calculated easily by performing a 
straightforward transformation which is 
detailed in the filter appendix. It may 
be shown that, at a single frequency, a 
given series R-C combination may be 
replaced with an equivalent parallel one. 
The input voltage generator is also 
replaced by a current generator. The 
resulting circuit is shown in Fig. 5B. 

The resistance across the resonator is 
now the parallel equivalent of R s \ R u 
and /?/,'. If this circuit is analyzed with 
respect to the loaded and unloaded Q of 
the resonator, it may be shown that the 
insertion loss of the resonator is given 
by 


1L = -10 log (1 - QlIQu) 2 (Eq. 10) 


In order to minimize the insertion loss 
of the filter, the loaded Q must be small 
in comparison with Q u . Noting the 
relation between resonator Q and its 3- 
dB bandwidth, this means that the 
bandwidth should be fairly large in 
order to hold the insertion loss down to 
a reasonable level. 

This characteristic is qualitatively 
true for much more sophisticated filters. 
However, the simple relationship of Eq. 
10 no longer applies with filters of more 
than one resonator.. 

Fig. 6 shows a general example of a 
multiple-resonator filter. In this case a 



( 6 ) 


Fig. 5 - Example of a filter which has loss. 

Advanced Receiver Concepts 115 








Fig. 6 - Example of a multiresonator filter. 


3-section filler is shown, although the 
general circuit configuration may be 
extended arbitrarily to any number. 

Capacitors are used in the 3-pole 
example of Fig. 6 in order to couple 
energy between the resonators, and to 
couple the source and load into and out 
of the filter. Inductive coupling could 
also be used, or a mixture of the two 
methods could be employed. 

The techniques of modern filter 
synthesis tell us that a given filter may 
be realized with resonators of equal Q„ 
if we establish the coupling between 
sections and control the singly loaded Q 
of the end sections. By singly loaded Q. 
we mean the loaded Q of the end 
resonator, when terminated, but with 
no coupling to the rest of the filter. 

Virtually any type of passband shape 
may be specified. Some of the common 
types include the Butterworth, 
Chebyshev and Gaussian responses. 
These names are ones that we often hear 
in connection with filters, but are rarely 
explained in the amateur literature. 
They are essentially mathematical terms 
naming the sometimes fairly compli- 
cated polynomials that describe the 
position of the poles of the filter in the 
complex frequency plane. In more prac- 


tion of a Butterworth filter is given by 


Atten (d B ) = 10 log (1 + S 2n ) 

(Eq. 11) 


where n is the number of resonators. S 
is the ratio given by 


_ f-fc 
-fc 


or 


_ fe- 
te ~h 


(Eq. 12) 


where/is the frequency of interest, f c is 
the center frequency of the filter, and 
f 3 + and / 3 _ are the upper and lower 
3-dB attenuation frequencies of the 
filter. Which form of the equation is 
used will depend upon whether the 
frequency of interest is above or below 
the center frequency of the filter. 

As an example of this equation see 
Fig. 7, where responses for a number of 
Butterworth filters are given. They are 


all designed for 3-dB attenuation fre- 
quencies of7.0and7.2 MHz. Curves are 
plotted for one through five resonators. 
The difference in skirt response as the 
number of tuned circuits in the filter is 
increased is profound, but there is a 
price to be paid. As the number of 
resonators is increased, the insertion loss 
will also increase dramatically for filters 
with a fixed bandwidth, all using the 
same type of resonator (constant Q u ). 

This is not the only effect of the loss 
elements in a filter. It turns out that the 
finite Q of the resonators complicates 
tire design. If classic image-parameter 
methods were used for the filter design, 
we would find that the filter shape 
would be distorted over that predicted 
when it was built and measured. In 
order to compensate for this effect, 
so-called predistorted filter tables (see 
the reference by Zverev in the bibliog- 
raphy) were used for the designs. 
Because of the subtlety, a general equa- 
tion set cannot be specified for the 
design. Furthermore, the filters f 
described in the appendix can not be 
scaled to other frequencies in the simple 
way that image-parameter filters can. 

As mentioned earlier, there is some- 
times an advantage to the use of capac- 
itive or inductive coupling over the 
other. When capacitive coupling is used, 
tire skirt response tends to be a bit 
steeper on the low frequency side. This 
is because the filter tends to degenerate 
into a high-pass structure away from the 
passband. Similarly, inductive coupling | 
seems to make the high-frequency skirt 
steeper. These effects become signifi- 



tical terms, they also lead to different 
filter characteristic shapes. The Butter- 
worth filter is one that is relatively flat 
across the passband. Indeed, this filter is 
often called a maximally flat response 
(mathematically, the first derivative of 
the transfer function vanishes at the 
center of the passband). The Chebyshev 
filter is somewhat more complicated. 

Some passband ripple may exist, but the 
skirt response close to the edges of the 
passband is steeper. The Gaussian 
response is not as fiat across the pass- 
band as the Butterworth or some 
Chebyshev filters. However, it has the 
advantage that "ringing” is minimal. 

Hence, Gaussian transfer functions are 
optimal for very narrow-bandwidth 
crystal filters, as an example. 

The filters described in the appendix 
are all designed for a Butterworth 
response. The main reason for this is 
that a Butterworth filter is among the 
easiest to align without resorting to 
advanced alignment techniques or 
extensive instrumentation. Hie attenua- Fig. 7 - Response curves for a number of Butterworth filters. 



li- 

re 

al 

te 


f 


H 


\ 


t\ 

ir 

F 

o 


fi 


It 

a 

Vi 

c 

e 

n 

a 

v 

n 


1 


116 Chapter 6 




cant well down on the response curves. 
For a 3-pole filter the differences 
become apparent when attenuations of 
more than 50 or 60 dB are achieved. 

If a narrow filter is designed so it 
may be tuned over a range of fre- 
quencies from the front panel of a 
receiver, proper coupling techniques 
should be used. If a multisection vari- 
able capacitor is used, inductive 
coupling is preferred between resona- 
tors. On the other hand, if a number of 
inductors are tuned simultaneously, 
capacitive coupling is desired. 

Although there are some exceptions, 
most filters using a multiplicity of 
resonators must be terminated properly 
at each end. The filters described in the 
appendix have components listed for 
termination of each end in 50 ohms. It 
is possible, however, to terminate them 
in much different impedances. The 
methods for achieving this are also 
outlined. 

A preselector filter that has become 
popular recently is tire so-called Cohn 
filter, This circuit is tunable from the 
front panel over a reasonable frequency 
range. The unusual characteristic of this 
circuit is that four resonators are used. 
However, only a three-section variable 
capacitor is required to tune it. The 
filter, as originally designed, was opti- 
mized for minimum loss in tire pass- 
band, making it ideal for receiver appli- 
cations. A representative circuit for the 
Cohn filter is given in Fig. 8. Generally, 
this circuit may be scaled to other 
frequencies. The 3-dB bandwidth may 
be increased by making the coupling 
inductors (1.45-jrH units in Fig. 8) 
larger in value. The skirt response can be 
made steeper by increasing the value of 
the shunt capacitors (270-pF units of 
Fig. 8). 

Mixer Design 

At the start of this chapter were 
concepts to define and measure the 
two-tone dynamic range of a receiver. 
The effects of adding or subtracting gain 
in a receiving system were discussed. 
However, little was said about the main 
origins of the 1MD which limits dynamic 
range. This topic is treated now. 

In the current state of the art we 
find that the design of filters and 
amplifiers is highly refined. By proper 
choice and application of transistors, 
low noise figure and high-intercept 
amplifiers are possible. The next section 
will present some of this information. 
Generally the mixer is the limiting 
element in a receiving system. If better 
mixers can be built, the amplifiers that 
are needed to accompany them are 
within reach, although still difficult to 
realize. 

An amplifier is a device dial relies 
upon the linear characteristics of a 
transistor in order to provide gain. By 



Fig. 8 - Tunable Cohn type of filter for 
1.8 MHz. L5 and L6 are 1 .45-pH bottom- 
coupling toroidal inductors. 

LI. L4 - 70 pH 
L2. L3 - 140 pH 


using devices that operate at high cur- 
rent levels, and by the application of 
feedback, this linearity can be 
emphasized. Similarly, filters employ 
passive elements which tend to be in- 
herently linear. However, in order to 
achieve mixing action, nonlinear opera- 
tion is desired. We must utilize square- 
law characteristics or the switching 
action in order to realize mixing. (The 
fundamental mathematics are outlined 
in die appendix.) Hence, in a device 
operated purposefully in a nonlinear 
mode, we would expect other responses, 
including unwanted ones, to occur. 

There are a number of devices that 
will function well as mixers. They all 
have dreir assets and problems. Some of 
these will be presented with some guide- 
lines for their use. 

The Dual-Gate MOSFET 

A popular mixer device in amateur 
equipment today, both commercially 
manufactured and homemade, is the 
dual-gate MOSFET. There are many 
varieties available. Unfortunately, ade- 
quate data are not provided by the 
venders, making it hard to say which is 
an optimum choice. Experiments sug- 
gest that the variations are not great. 

There is good reason for the popular- 
ity of the MOSFET. It is a device that 
can provide considerable gain (some- 
times desired). Furthermore, the noise 
figure is fairly low and the output 
intercept is rather high, especially when 
minimum power consumption is con- 
sidered. Finally, the local-oscillator 
power required is low, making the de- 
vice easy to apply. 

A typical mixer is shown in Fig. 9. 
In til is circuit pi networks are used to 
match both the input (gate I) and the 
output at the drain. This is done to 
establish the impedances seen at the two 
ports of the device. A variable voltage 
bias source is used to establish the 
operating conditions at gate 2 which 


lead to the best performance. The out- 
put was applied to a spectrum analyzer 
while the input was driven from a pair 
of signal generators which were added in 
a hybrid combiner. An attenuator was 
used after the combiner in order to 
ensure proper operation of that compo- 
nent. (An easily made combiner will be 
described later for use in the amateur 
shop.) A third generator was used as an 
LO. 

First, it was found that the gain of 
the mixer was dependent upon the 
terminating impedances and the level of 
the LO voltage applied to gate 2. There 
was also some variation when other 
similar device types were used in the 
circuit. Of major significance is the fact 
that the conversion gain was always 
about 12 dB lower than the gain of the 
same device when operated as an ampli- 
fier with the same termination imped- 
ances. This implies that the conversion 
transconductance is 1/4 of that dis- 
played when the same device is operated 
as an amplifier. This optimum gain 
occurred with an L.O injection of about 
5 volts pk-pk at gate 2. It was also 
found that die optimum dc bias voltage 
for gate 2 was about 1 volt. This tells us 
that the common practice of attaching 
gate 2 to the source of the device 
through a large resistor is a good one. 

The intermodulation distortion per- 
formance was good. With a 2000-ohm 
termination on the drain (at 9 MHz) the 
output intercept for third-order 1M was 
+ 19 dBm. This same output intercept 
was obtained when the device was oper- 
ated as an amplifier at 14 MHz (same 
termination impedances). When the 
MOSFET was operated as an amplifier 
or a mixer, gain compression occurred 
just a few dB below this intercept level. 
The 5-volt pk-pk LO injection appeared 
optimum for both blocking and IMD 
performance. 

The nature of the output termina- 
tion is critical with this mixer. In the 
experiment outlined, die output of the 
pi network at the drain was the 50-ohm 
input of a spectrum analyzer. This 
termination was quite flat at virtually all 
frequencies. This is not typical in the 
usual application. The more common 
terminadon for the mixer is the input of 
a crystal filter. While the filter may 
appear to be a clean resistive termina- 
tion within the passband of the filter, 
the input impedance is usually quite 
different at otliei frequencies. The usual 
ladder type of filter looks something 
like an open circuit at frequencies near 
(but not exactly in) the passband of the 
filter. If this were applied direedy to the 
drain of the mixer, the results could be 
quite compromising. The reason is that 
a signal which can cause undesired 
distortion effects is usually not the 
signal to which the receiver is tuned. 
Hence, when this signal is heterodyned 


Advanced Receiver Concepts 117 







in the mixer, the output will not lie 
within the passband of the filter. This 
can result in large voltage excursions at 
the drain, leading to blocking or IMD. 

The pi network used in Fig. 9 is one 
of the better choices as a matching 
mechanism to work into a crystal filter. 
The reason for this is that die pi 
network has an impedance-inversion 
property. That is, if the output termina- 
tion is less than that for which it was 
designed, the input impedance appears 
higher than the design center. On the 
other hand, if the output termination 
appears high in imoedance value, the 
input seen at the drain is low. The latter 
situation is desired. When die input 
impedance of the crystal filter appears 
to be an open circuit (out of the 
passband), the load presented to the 
drain approaches that of a short circuit. 
This prevents large voltage excursions. 

Sabin suggested the use of another 
type of impedance inverting network 
( QST , July, 1970). lie used an under- 
coupled double-tuned circuit. This kind 
of network has the advantage that it 
acts as a bandpass filter. This protects 
the crystal filter and following circuits 
from spurious filter responses dial 
sometimes occur. 

There is another mixer output that 
might be investigated as a possible 
source of IMD - the image. In the 
circuit of Fig. 9, the LO frequency is 23 
MHz, and the input is at 14 MHz. The 
desired i-f is 9 MHz. However, the mixer 
will produce not only difference fre- 
quencies (23 14) but sums also, in this 

case at 14 + 23 = 37 MHz. It is possible 
that the existence of these currents in 
the drain would degrade the output 
intercept. No experiments were per- 
formed to achieve a proper termination 
for this frequency. 

There is a problem with pi-network 
matching that has not been mentioned. 
Although the network has the advantage 
of presenting a proper load to the drain 
of die MOSFET in order to minimize 
blocking, it docs not provide an output 
that terminates a filler properly. The 
output impedance of the FET is much 


higher dian the 2000-ohm value for 
which our network was designed. It may 
be as high as 100 kS2. If the pi network 
was designed for a value this high, the 
conversion gain would be very high, but 
die output intercept and blocking level 
would be degraded severely. As a result 
of the need for filter termination, it is 
common practice to put a resistor 
widiin the output-matching section. 
This resistor will absorb part of the 
available output power, with degrada- 
don of the output intercept as well as 
reduced gain. 

Detailed noise-figure measurements 
were not performed with the test circuit 
of Fig. 9. However, in testing a number 
of receivers with dual-gate MOSFET 
mixer front ends, with low-loss input 
matching, we found that noise figures of 
8 to 10 dB are common. Careful design 
may improve this. 

The detailed performance evalua- 
tions just presented may sound pessi- 
mistic. However, this is not die case. 


The results are quite acceptable, espe- 
cially when the ease of application of 
the MOSFET is considered. Some 
single-conversion receivers using such a 
front end were evaluated. TTiey dis- 
played a two-tone dynamic range of 
over 90 dB, which is better dian most 
commercially available units. Receivers 
with poorly applied MOSFET mixers 
often have a DR as low as 60 or 70 dB. 

Diode Mixers 

Next to the dual-gate MOSFET, the 
most common mixers in amateur re- 
ceivers are those using diodes. This class 
has a number of advantages. The first 
one is that they are inherendy broad- 
band. Therefore, they are applied easily 
to multiband designs. Another ad- 
vantage is the relatively low noise figure. 
Most diode mixers generate very litde 
noise. As a result die noise figure is 
nearly the conversion loss of the mixer. 
Another asset is that diode mixers dis- 
play high intercept points. Finally, most 
diode mixers arc balanced. The implica- 
tions here are twofold. First, die bal- 
ance has the effect of preventing energy 
applied to die LO port of die mixer 
from appearing at the i-f or rf ports. 
Second, certain types of noise (a-m 
noise) that would appear at the LO port 
all attenuated when diey reach the i-f 
port, even if diat noise might actually 
be at the i-f. Balance can also improve 
IMD immunity. 

In spite of die virtues, diode mixers 
have their faults. They require high LO 
power in order to provide optimum 
performance. Proper termination of the 
mixers is critical, especially at die i-f 
port. Finally, depending upon diode 




118 Chapter 6 


Fig. 10 - Circuit of a doubly balanced diode mixer. The diodes are HP-2800S. 








type, many mixers of this kind are 
prone to harmonic mixing. This phe- 
nomenon was discussed in connection 
with diode product detectors (chapter 

A double-balanced diode-ring mixer 
is shown in Fig. 10A. The usual mixer 
of this type contains hot-carrier diodes, 
although high-speed silicon switching 
diodes are used sometimes. The most 
critical detail in building a mixer of this 
kind is in the winding of the trans- 
formers. The characteristics of the trans- 
formers will be the main factor that 
limits the bandwidth of the mixer. The 
balance (die ratio of the power at one 
port which appears at one of the others) 
will depend upon the transformer 
quality and upon the uniformity of the 
diodes, 

If a diode-ring mixer is built to cover 
the hf spectrum, the transformers 
should be wound on high permeability 
ferrite toroids. A typical transformer 
would contain 10 trifilar turns of No. 
30 enameled wire on an Amidon 
FT-37- 43 core. It is useful to employ 
wires of three different colors. If this is 
not possible, care should be used to 
ensure that the proper windings are 
identified: The section in chapter 4 on 
transformer design should be consulted. 

If a mixer is built to cover the vhf or 
lower uhf spectrum, cores with low 
permeability are often used. A typical 
value might be 125 (Q1 material, or 
Amidon type-61). Toroids do not al- 
ways present the optimum geometry for 
such applications. Excellent mixer trans- 
formers can be built using ferrite beads 
with multiple holes. 

In applications where good balance 
is desired over a very wide bandwidth, it 
is useful to add another transformer or 
two. This is realized by driving each 
balanced port with an isolating “sorta- 
balun." This scheme is shown in Fig. 
10B. Balance of 60 dB or more in the hf 
region is not unusual. 

Several mixers of the simple ring 
configuration (Fig. I0A) have been in- 
vestigated experimentally. These in- 
cluded homemade mixers and com- 
mercially available units. There is no 
significant difference between the two 
except in cases where extremes of bal- 
ance or bandwidth are desired. 

When using llP-2800 hot-carrier 
diodes and transformers like those just 
described, the typical conversion loss is 
6 to 7 dB. This value is constant over 
most of the mixer bandwidth, reaching 
higher levels at very high and very low 
frequencies. Although the signal- 
handling capability of each mixer will 
differ, a good rule of thumb is that the 
output intercept of simple rings is 
roughly equal to the level of LO power 
applied. This is extremely important in 
the design of wide-dynamic-range re- 
ceivers. Most diode mixers will achieve 


close to minimum conversion loss with 
as little as one or two milliwatts of LO 
power. However, for best IM perfor- 
mance, it is wise to increase the LO 
power to +10 to +13 dBm. or even more 
if die diodes will handle the larger 
currents. 

The measurements of output inter- 
cept outlined in the foregoing were 
obtained with a test setup like that used 
for the evaluation of the dual-gate 
MOSFET, with the i-f port terminated 
in the 50-ohm input of a spectrum 
analyzer. A common result with simple 
ring mixers was an output intercept of 
+15 dBm with an LO power of +13 
dBm. After the initial measurements 
were performed with broadband ter- 
minations at all ports, tuned circuits 
were inserted in various lines to the 
mixer. These were single-tuned LC cir- 
cuits. The results were profound! When 
a single tuned circuit was put in the i-f 
port it had the effect of still presenting 
a 50-ohm termination at the desired i-f 
of 9 MHz. (The rf energy was at 14 MHz 
and the LO was at 23 MHz.) However, 
at frequencies other than the 9-MHz i-f, 
die impedance seen was highly reactive. 
This had the effect of decreasing the 
output intercept from +15 dBm to +5 
dBm in several of the mixers studied. 
The conversion loss did not change 
significantly. 

When a narrow-band termination 
was used at the rf and LO ports of the 
mixer, a degradation in output intercept 
was also observed. However, it was not 
nearly as severe as that seen at the i-f 
port. 

The critical frequency that must also 
be terminated in the diode mixer is the 
image. In die case outlined, this would 
be die sum of the rf and LO fre- 
quencies, or 37 MHz. If this energy is 
not absorbed in a resisdve termination, 
it may be redected back into the ring 
where it can interact with existing sig- 
nals to produce IMD. 

There are two general approaches to 
this terminadon problem. One is 
through the use of attenuators. A 3- or 
6-dB pad is often used at the output of 
the mixer to ensure a broadband ter- 
mination. Unfortunately, this attenua- 
don adds directly to the noise figure of 
the mixer. A more satisfactory solution 
is to terminate the i-f port in a diplexer. 

A diplexer is a network of resonant 
circuits that is arranged so that die 
desired frequency is passed through the 
network with minimal attenuation. 
However, additional inductors and ca- 
pacitors are arranged so that other 
frequencies are terminated also. That is, 
the network has an input impedance 
which is close to 50 ohms at all the 
frequencies of interest. Two possible 
configurations are presented in Fig. 11. 
The first is a combination of bandpass 
filters. Cl, C2, C3 and LI form the 


single-pole bandpass filter. At fre- 
quencies other than the 9-MHz design 
center of the filter, the input impedance 
will be capacitive. The out-of-band 
energy is handled by Rl, C4 and L2. 
The inductor and capacitor are also 
resonant at 9 MHz. At the i-f frequency 
tiiey appear as a high impedance. Mini- 
mal current flows in Rl. When the 
frequency departs from 9 MHz con- 
siderably, L2 and C4 appear as a Icw- 
impedance path to ground. Now Rl is 
directly across die mixer output, pro- 
viding a proper termination. 



Fig. 1 1 — Diplexer circuits for use after a 
mixer. 


The other diplexer shown uses a 
combination of a low-pass and a high- 
pass filter. The circuit operates in a 
similar fashion to the bandpass design 
just described and is especially useful in 
receivers using low intermediate fre- 
quencies (such as 455 kHz). The low- 
pass filter is a T network cut for the 
i-f of interest. The high-pass filter 
should be designed for a 50-ohm charac- 
teristic impedance and a cutoff fre- 
quency of about three times that of the 
i-f. Such a filter has reactances equal to 
the characteristic impedance at the cut- 
off frequency. 

Some measurements of noise figure 
and IMD suggest that the termination of 
a diode-ring mixer at the i-f may not be 
as critical as the image termination. This 
leads to the possibility of accepting 
some compromise in match at the i-f in 


Advanced Receiver Concepts 119 





Fig. 12 — Examples of multi-diode high-level ring and dual-bridge high-level mixers. 


order to obtain improved system noise 
figure. An example would be a dual-gate 
MOSFET low-noise amplifier following 
the mixer. In this case it would still be 
necessary to provide proper termination 
for the image energy, still making a 
diplexer desirable. In stringent designs, 
all products resulting from harmonic 
mixing should be terminated. Such con- 
siderations emphasize the need for do- 
ing broadband designs with good match- 
ing well into the vhf spectrum, even 
when the receiver is for use on the hf 
bands. 

There are other diode mixers that 
offer improved intercept characteristics 
with virtually no compromise in noise 
figure. Two of these are shown in Fig. 
12. In the first, the single diodes have 
been replaced by a series combination 
of two diodes. This helps the mixer to 
accept higher LO power without burn- 
ing out the diodes. In the ring configura- 
tion, even when multiple diodes are 
used, the limitation is the maximum 
current that the diode can handle. Re- 
verse voltage breakdown is not a prob- 
lem, since each pair of conducting 
diodes protects the reverse-biased ones. 
In multidiode mixers (more than 4) 
designed for high intercept factors, care 


should be taken to ensure that the 
diodes are well matched. Diodes with a 
high junction area are desired also, since 
they will handle larger currents. 

The second mixer (Fig. 12) departs 
from the ring configuration: A pair of 
bridge rectifiers is used. The local oscil- 
lator is applied to each bridge in paral- 
lel. However, the bridges are arranged 
with respect to the LO transformer so 
that only one is “on" at one time. Each 
bridge conducts on alternating half 
cycles of the LO waveform. The bridge 
that is on at any instant connects that 
end of the rf-port transformer to 
ground. The opposite side of the rf-port 
transformer is, in effect, connected 
directly to the i-f port. 

One unusual characteristic of the 
second mixer of Fig. 12 is that resistors 
appear in the local oscillator lines to 
each bridge. These resistors cause signifi- 
cant effects. They allow the LO port to 
be driven with higher voltages than 
would be possible otherwise. This not 
only leads to high currents flowing in 
the diodes during their “on" half cycle, 
but it allows a larger reverse voltage to 
be established across the "off’ bridge. 
This causes the diodes to operate in 
nearly a true switching mode. Note that 


the resistors are not in the rf to i-f path. 

The classic diode ring is analyzed 
best if die diodes are thought of as 
switches that are controlled by the LO 
signal. In this condition, an incoming rf 
signal is “chopped” at the LO rate. A 
mathematical analysis will show that 
this leads to sum-and-difference fre- 
quencies. Detailed study indicates that 
tlie IMD effects which limit the inter- 
cept are a result of departures from the 
switching action. If a weak sine-wave 
drive is used at file LO port, the diodes 
will spend a portion of each cycle near a 
zero-bias condition. Because of this, 
strong rf signals can have a major effect 
in changing the conduction state of the 
diodes. On the oilier hand, if the mixer 
is driven with a much stronger LO, and 
ideally even a square wave, the diodes 
are allowed to spend a much shorter 
portion of each cycle near this zero-bias 
point. The stronger mixers are those 
that allow large LO signals to be 
applied, and permit larger reverse volt- 
ages to appear across nonconducting 
diodes. 

Both of the mixer types described 
have been studied by the writers. The 
original designers of these mixers are 
not known to the writers. Both have 
been outlined in recent papers (see the 
bibliography: Cheadle, 1973, and 
Rohde, 1975) although only the multi- 
diode ring is described in detail. Our 
measurement results were virtually iden- 
tical for the two mixers. The insertion 
loss was about 6.5 dB and the output 
intercept was +22 to +23 dBm. The 
frequencies were the same as those used 
in the other evaluations. It was found 
that the dual-bridge mixer exhibited 
extremely good balance, up to 60 dB in 
the hf region. 

One problem that was noted with 
both high-level mixers was that they are 
not always “well behaved.” This means 
that the intermodulation distortion 
products did not always drop by 3 dB 
when the input tones were decreased by 
1 dB. Although it is conjecture at this 
point, this departure could result from 
mismatch in the diodes at specific cur- 
rent levels, or from nonlinearities in the 
ferrite transformers. The intercepts 
quoted are indicative of the well- 
behaved range of operation with an LO 




Fig. 13 — A singly balanced mixer which uses 
two diodes. 


120 Chapter 6 




9MH* 



Fig. 14 - Circuit of a JFET mixer. 


drive of +17 dBm. 

All of the diode mixers discussed 
have been doubly balanced designs. 
That is, balanced transformers have 
been used at two of the three ports. 
However, it is not mandatory that a 
mixer be doubly balanced in order to 
assure strong performance. Shown in 
Fig. 13 is a singly balanced mixer using 
only two diodes. This design has the 
virtue that large voltages can be estab- 
lished across the diodes in the off 
condition. The center tap of the LO 
transformer is grounded. This improves 
balance. If this configuration is used, 
the i-f and rf are applied to the connec- 
tion of the diodes. A diplexer is used to 
isolate the two frequencies. At the 
lower frequencies it may be acceptable 
to extract the i-f from the center tap. 

One virtue of mixers of this kind is 
that they often have a lower insertion 
loss than is typical of the four-diode 
mixers. Such a mixer has an insertion 
loss of 5 dB with an output intercept of 
+15 dBm. These results were obtained 
with an LO drive of +15 dBm. Two- 
diode mixers are popular for vhf and 
uhf application. 

Mixers Using JFETs 

Some JFETs can provide exception- 
ally good performance as mixers. They 
are, however, more difficult to use than 
MOSFET mixers. 

Shown in Fig. 14 is a 2N4416 mixer. 
The properties are similar to those 
obtained with the dual-gate MOSFET: 
The input impedance is high and the 
conversion gain is commensurate with a 
transconductance of 1/4 that seen with 
the same device operated as an ampli- 
fier. Biasing is critical. It should be 
chosen so that the gate-to-source voltage 
is equal to 1/2 of the pinchoff voltage 
of the device. The local oscillator signal, 
which is applied to the source, should 
be as large as possible within the con- 
straints that the device should never go 
into the pinchoff region, nor should the 
gate diode be driven into conduction. 
This means that the pk-pk LO voltage 


should be a little below the pinchoff 
voltage of the FET. 

The MOSFET mixer had a high 
output impedance. On the other hand, a 
JFET has a lower value, typically 
around 10 kf2 for the resistive portion. 
This makes matching to filters a bit 
easier. An impedance inverting network 
should still be used. 

The major advantage of the JFET 
mixer over the MOSFET is that the 
noise figure is lower. Values as low as 4 
dB have been reported (Sabin, 1970). 
The writers have not done intercept 
measurements on this mixer. 

Shown in Fig. 1 5 is a mixer using a 
dual JFET (Siliconix E430) which has 
been designed especially for mixer appli- 
cations. The input transformers are sim- 
ilar to those used in diode mixers. Pi 
networks are used at each drain to do 
part of the impedance matching as well 
as perform impedance inversion. Each pi 
network is designed to transform from 
2000 ohms at the drains to 100 ohms. 
The push-pull 100-ohm outputs add to 
form a 200-ohm balanced source. This is 


transformed to a single-ended 50-ohm 
output with a trifilar transformer. The 
LO power requirement for this mixer is 
fairly high, since the sources are driven 
rather than the gates. With a +17-dBm 
LO drive, an output intercept of +26 
dBm was measured. The gain was 2 dB. 
Noise figures from 6 to 8 dB are quoted 
as typical by the manufacturer. 

Doubly balanced mixers using four 
JFETs have also been described. Al- 
though the writers have not investigated 
them (yet!), they appear to offer great 
promise. 

Mixer Comparisons 

Great care should be used when 
comparing mixer designs. Many workers 
suggest that mixer gain is an advantage. 
This is not necessarily true. Compare, 
for example, a dual-gate MOSFET mixer 
with a simple diode ring. The former 
may have a gain of 20 dB, an output 
intercept of + 18 dBm, and a noise figure 
of 10 dB. If a receiver is built with this 
mixer as the front end, driving a filter 
directly, the MDS will be -137 dBm. A 
bandwidth of 500 Hz was assumed. The 
input intercept of this receiver will be 
+18 dBm 20 dB. or -2 dBm. Recall- 
ing that DR = (2/3) (P, - MDS), the 
dynamic range of the system will be 90 
dB. 

Consider now the simple diode ring 
with a conversion loss of 6 dB. Assume 
that the circuit following the ring is 
strong, has a noise figure of 3 dB, and 
that a preselector filter with a 1 -dB loss 
is used ahead of the mixer. The overall 
system noise figure will again be 10 dB, 
leading to an identical MDS of -137 
dBm. The input intercept of the mixer 
will be tire output intercept plus the 
conversion loss. Assume that the output 
intercept is +1 5 dBm. The receiver input 
intercept will now be +15 dBm + 6 dB 



Fig. 15 — A high-level balanced JFET mixer. T1, T2 and T3 contain 10 bifilar turns of No. 28 
enam. wire on FT-37-61 toroid cores. T4 has 10 trifilar turns of No. 28 enam. wire on an 
FT-37-61 core. 


Advanced Receiver Concepts 121 






Pi- + 1dBm (C) 


Fig. 16 — Illustration ot a receiver which has no gain ahead ot the (liter (A). At (B) and (C), 20 dB 
of gain has been added. 


(mixer loss) + 1 dB (preselector loss) = 
+22 dBm. The dynamic range of the 
receiver is now 106 dB! In this case, the 
loss of the diode mixer is a profound 
advantage, leading to a 16-dB increase in 
dynamic range . 

For general applications in straight- 
forward receivers, the dual-gate 
MOSFET is highly recommended. For 
improved performance, simple diode 
mixers are suggested. However, more 
care is required in designing the cir- 
cuitry following tire mixer. For the 
experimentally inclined amateur with 
instrumentation for evaluation of the 
circuits, high-level mixers using diodes 
or balanced JFETs are suggested. The 
advanced amateur may build equipment 
to do this evaluation himself (see QST, 
July, 1975). 

Front-End Amplifiers 

There are two major ways in which 
amplifiers are used in the front-end 
section of a superheterodyne. Hie 
classic one is as an rf preamplifier 
preceding the mixer. The other, which is 
not quite as traditional, is as an i-f 
amplifier following the mixer. In multi- 
conversion systems, amplifiers are often 
used in between the mixers. 

Consider a single-conversion receiver 
designed for cw operation. Assume that 
the bandwidth of the crystal filter is 
500 Hz (27 dB above one Hz), and that 
a simple diode-ring mixer is used. The 
mixer will have a 6-dB insertion loss anu 


a +12-dBm output intercept. Assume 
that the noise figure of the i-f amplifier 
following the crystal filter is 5 dB. 

As a start, imagine a receiver that has 
no gain ahead of the filter. This system 
is shown in Fig. 16A where the prese- 
lector network is assumed to have a 
3-d B insertion loss. The crystal-filter 
loss is 4 dB. Also, we will use a 3-dB 
attenuator between the mixer and the 
crystal filter to ensure that the output 
intercept of the mixer is preserved. 
Using the methods outlined in the 
earlier sections of this chapter, this 
system can be analyzed. The results are 
noise figure = 2 1 dB. MDS = - 126 dBm. 
P, (input intercept) = +21 dBm and DR 
= 98 dB. 

Consider now the modified receiver 
shown in Fig. I6B. Here a very strong 
amplifier with an output intercept of 
+40 dBm and a gain of 20 dB is inserted 
between the mixer and the crystal filter. 
The input intercept of this amplifier will 
be +20 dBm. Since this is quite a bit 
greater than the output intercept of the 
mixer that drives the amplifier, we will 
assume that the amplifier is virtually 
free of IMD. A noise figure of 4 dB is 
assumed for the amplifier. Analysis of 
this design gives the following results: 
NF = 13 dB, MDS = 134 dBm. F, = 

+21 dBm, and DR = 1 03.3 dB. We have 
gained 8 dB in sensitivity and about 5 
dB in overall dynamic range, while 
leaving the input intercept constant. 

The third case for consideration is 


shown in Fig. I6C. Here the same 
amplifier has been placed as an rf 
preamplifier between the preselector 
and the mixer. Analysis yields noise 
figure = 8 dB. MDS = 139 dBm, F, = 

+ 1 dBm, and DR = 93.3 dB. The low 
noise gain has yielded an improvement 
in noise figure, but has brought about a 
dramatic decrease in input intercept and 
dynamic range. For most amateur appli- 
cations, case B would be the optimum. 
Clearly, gain distribution is a vital con- 
sideration. 

The criteria for the design of pre- 
amplifiers and post-mixer amplifiers 
differ considerably. In the case of die 
latter, the amplifier input intercept 
should exceed the output intercept of 
the mixer used. For the preamplifier, I 
the output intercept should exceed the 
input intercept of the mixer. If these 
criteria are not met, IMD from the 
amplifiers will add to that generated 
within the mixer. 

Post-Mixer I-F Amplifiers 

Amplifiers operating at Ure inter- 
mediate frequency, and following the 
mixer directly, should have output in- 
tercepts of +30 dBm or more, In search- 
ing the literature we find that common- 
ly available FETs.both die junction and 
MOS types, are not strong enough. This 
leaves the job to bipolar transistors. FET 
technology is changing rapidly, however, 
and there are indications that much 
better units will be available in the future. 

Shown in Fig. 17 is an amplifier that 
was breadboarded for preliminary inves- 
tigation. This amplifier used an Amprex 
BFR-94 transistor. This is a stud- 
mounted power device designed for 
cable-TV applications. No impedance 
matching was performed at either the | 
input or die output. Still, at 10 MHz the 
transducer gain was well over 25 dB and 
die noise figure was about 5 dB. The 
output intercept of this amplifier was 




Fig. 17 — A post-mixer amplifier without ' 

feedback. j S 


122 Chapter 6 




+40 dBm. The feature of this circuit is 
that there was 100 mA of collector 
current (lowing in the transistor. At this 
level, the saturated power output of the 
amplifier was over one -half watt! 

After the initial experiment, the 
BFR-94 circuit was modified. A 2:1 
turns-ratio transformer was placed in 
the collector circuit, providing a 200- 
ohm collector load resistance. Also, 
negative shunt feedback was introduced 
by a 1000-ohm resistor, ac coupled 
from collector to base. With this modifi- 
cation the output intercept went up to 
+45 dBm. Noise figure was not mea- 
sured. Input matching would be required 
when using the modified circuit, for 
shunt feedback will have the effect of 
depressing the input impedance well 
below 50 ohms. 

In general, past amplifiers made 
from bipolar transistors will use negative 
feedback as well as some impedance 
matching at the output. An amplifier 
from one of the writers’ receivers is 
shown in Fig. 18. A Nippon Electric 
2SC-1252 transistor is biased to 65 mA 
of collector current. A 2: 1 turns-ratio 
ferrite transformer was used at the 
output, presenting a load of 200 ohms 
to the collector. Emitter degeneration 
and shunt feedback were employed. 
This combination has the result of 
controlling stage gain as well as the 
input and output impedances. Without 
the 6-dB attenuator in the output, the 
amplifier provided 23 dB of gain and an 
output intercept of +41 dBm. The noise 
figure was 6 dB at 10 MHz, and the 
input match to 50 ohms was excellent 
(30-dB return loss over the hf spec- 
trum). NEC transistors are available 
from California Eastern Labs of Burlin- 
game, California. 

A 6-dB attenuator is included in the 
output in actual application. This has 
the effect of reducing the net gain to 17 
dB and dropping the output intercept to 
+35 dBm. However, it has the asset of 
keeping the input impedance of the 
amplifier relatively constant at all fre- 
quencies. If it were not there, variations 
in input impedance of the crystal filter 
that follows the amplifier would reflect 
back through the amplifier to cause 
variations in the input impedance. This 
characteristic is typical of amplifiers 
with heavy shunt feedback. Additional 
information on the design of negative- 
feedback Class A amplifiers.is presented 
in connection with our discussion of ssb 
methods. 

The NEC transistor was a convenient 
unit to use. It is mounted in a TO-5 
package. However, unlike most TO-5 
devices, the collector is not common to 
the case. There is good internal thermal 
bonding, nonetheless. In our application 
a suitable hole was drilled in the circuit 
board allowing the transistor to be 
soldered to the ground foil: The board 


serves also as a heat sink. 

A general equation may be applied 
to bipolar transistors to estimate their 
output-intercept characteristics. It is 
assumed that the collector is terminated 
in a 50-ohm load. Under these condi- 
tions, the output powers in dBm for 1 
dB of gain compression, and for IM 
intercept, are given by 

/’(compression) = - 16 + 20 log IO / c 

P„ = 20 log I0 / c = output intercept 

(Eq. 13) 

where / t . is the collector current in mA. 
These equations should be regarded as 
an optimistic rule of thumb rather than 
as an absolute definition of the perfor- 
mance. The intercept may often be 
improved by impedance matching to the 
collector. This was the case in the 
2SC-1252 amplifier. Deriving the 
equation for gain compression is 
straightforward: The output power is 
that where the peak signal current is 
equal to the standing dc current. It is 
surprising to the writers that these 
simple relationships are so accurate in 
practice. 

There are some general requirements 
for the choice of transistor types for 
amplifiers of this kind. From the 
.equations we see that a high output 
intercept will result only from a high 
collector current in the transistor. The 
transistor must be capable of operating 
at high currents and of dissipating tire 
power. However, a reasonably low noise 
figure is also desired. Usually, feedback 
needs to be applied. Because of these 
criteria, the transistor should have a 
very high f r . In the two circuits presented 
die devices have gain-bandwidth prod- 
ucts of well over 1 GHz. For applica- 
tions with I e of approximately 100 mA. 



Fig. 18 — A bipolar type of post-mixer ampli- 
fier which uses feedback. 


the Amperex BFR-94 and A-209 types, 
the NEC 2SC-1252 as well as die Moto- 
rola 2N5947, are suggested. For ampli- 
fiers with up to 50 mA, the Amperex 
A-210, Motorola 2N5943 or RCA 
2N5109 are suggested. Vhf power tran- 
sistors are worth consideration. Exam- 
ples would include die 2N3553 and 
2N3866. For strong bipolar amplifiers 
in the vhf and uhf region, the NEC 
V021 is recommended. With l c = 30 
mA, this device will give an 18-dB gain 
and 4-dB noise figure at 432 MHz, 
without careful matching. 

Preamplifier Design 

The criteria for the design of ampli- 
fiers that precede the mixer in a super- 
heterodyne are somewhat different than 
those for post amplifiers. First, the 
intercept requirements are not as strin- 
gent. Since the usual diode-ring mixer 
will have an input intercept of +15 to 
+ 18 dBm, amplifiers only a bit stronger 
than this will suffice. Second, lower 
noise figures are usually desired. Both 
FETs and bipolar transistors may be 
used. 

FETs have some general advantages. 
Less current is required in order to 
realize an equivalent output intercept. 
Their noise figures are quite low in the 
hf region. Finally, their output powers 
for gain compression are closer to the 
jutput intercept than is the case for 
bipolar transistors. This means they are 
less prone to blocking problems. 

In spite of the virtues of FETs, 
bipolar transistors may be used quite 
successfully as hf preamplifiers. They 
come into their own in the vhf and 
microwave regions. The major advantage 
of the bipolar transistor over the FET is 
that it has well defined input and 
output impedances and is much more 
easily used with negative-feedback sys- 
tems. This can be of profound impor- 
tance if a low-loss preselector is used 
ahead of such an amplifier. If pre- 
selector performance is to be main- 
tained, the filter must be terminated 
properly. In the hf region this is not 
possible with FETs operating in the 
common-source configuration. A clean 
input match is realized with an FET only 
if a resistor is added for termination. 
This has the effect of degrading the gain 
and noise figure. This compromise may 
be altered with the application of ad- 
vanced feedback methods. 

Although the theory is beyond the 
scope of this text, it is possible to apply 
advanced feedback methods to bipolar 
transistors to great advantage. The re- 
sults are that low noise figure and a 
good input and output match may be 
obtained simultaneously. One of our 
colleagues (WA7TZY) has built ampli- 
fiers using bipolar transistors with equiv- 
alent noise temperature under 100°Kat 
432 MHz, with input and output return 


Advanced Receiver Concepts 123 



is still possible. Noise figures of just over 
1 dB have been reported for such 
amplifiers in the hf region. 

A common-gate JFET amplifier is 
shown in Fig. 21 . It is claimed that such 
a circuit is inherently stable. This is not ] 
necessarily true, as can be demonstrated 
with a stability analysis using two-port 
network theory (see the appendix for 
comments on stability analysis). The 
spurious oscillations that might occur 
with the common-gate circuit are 
usually in the vhf or uhf region and are 
often cured with a small resistor in 
series with the drain. With clean circuit 
layout, instabilities in the hf region are 
rarely a problem. The noise figure of 
Utis circuit can be close to that of the 
same device operated in the common- 
source configuration. The available 
power gain is not as high, with values of 
10 to 14 dB being typical. An advantage 
of the common-gate circuit is that the 
input impedance is well defined and 
losses of better than 20 dB. In general, the advantage that it operates over a fairly lav. It is approximated by R in = 

the simple resistive feedback methods wide band of frequencies. The first \/g „, , where g m is the common-source 

shown for post amplifiers (Fig. 18) have amplifier, which uses a coil from gate to transconductance. For devices like the 

tiie effect of degrading the noise figure, drain, provides cancellation of the effect 2N44I6 with g„, near 5000 micromho, 

(See the analysis in the appendix.) of the gate-to-drain capacitance only at a 200-ohm input is produced This is 

An excellent choice for general- one frequency. Oscillation at fre- easily matched to 50 ohms by means of 

purpose bipolar amplifiers in the hf quencies outside the band of operation a 2:1 turns ratio ferrite transformer, 

region is the 2N5179, biased to approxi- 
mately 20 mA. The Amprex BFR-91, 
biased between 10 and 20 mA, is 
excellent for the 144- and 432 -MHz 
bands. 

In spite of the input-match problem 
with FETs, they can have low noise 
figures. Shown in Fig. 19 is a preampli- 
fier using a 40673 dual-gate MOSFET. 

A pi network is used for input 
matching, transforming the input 50- 
ohm source to an impedance at gate 1 
between 2000 and 3000 ohms. The 
loaded Q of the network should be as 
low as possible if minimum noise figure 
is desired. Several hf amplifiers built by 
the writers had noise figures under 2 dB. 

The MOSFET amplifier should have 
careful bypassing at gite 2. The capac- 
itor should be effective up to I GHz. 

Otherwise, drain-voltage vaiiations will 
couple back through gate 2 to the input. 

That can cause oscillations in the lower 
uhf spectrum. In one amplifier built for 
14 MHz, an oscillation was found at 800 
MHz. It was cured by placing a 470-pF 
capacitor in parallel with the existing 
.01 -jrF one, and by reducing the pigtails 
of the FET as much as possible. Reisert 
(W1JAA) has solved this problem by 
placing a ferrite bead on the gate-2 lead. 

He reported noise figures of under 1 dB 
with circuits like the one of Fig. 19, 
using a 40673 operated at 28 MHz 
(Ham Radio. Oct. 1975). 

Shown in Fig. 20 is a pair of am- 
plifiers using JFETs which arc oper- 
ated in the common-source configura- 
tion. Neutralization is used to stabilize 

the amplifier. Bridge neutralization has Fig. 20 - A pair of JFET amplifiers which operate in the common-source mode. 




Fig. 19 - A low-noise preamplifier using a dual-gate MOSFET. Z1 and Z2 are pi networks with 
Q values of 10 or less (see text). 



1 24 Chapter 6 




Fi 9- 21 ~ C' r «*it of 3 common-gate JFET amplifier. Typical gain is 10 dB, and output inter- 
rept is +26 dBm. Select R to provide a low-input VSWR. T1 contains 10 bifilar turns of wire 
on an FT-37-61 toroid core. 


This would provide a good broadband 
termination for a preselector network. 
A good input match here would prob- 
ably degrade noise figure. 

The major point to emphasize when 
considering preamplifiers for hf re- 
ceivers is that the gain must be chosen 
carefully. Excess gain will do little to 
improve noise figure beyond the value 
that is needed. However, it can have 
disastrous effects on the overall dy- 
namic range of the receiver. 

Oscillators for Receiver Application 

The problems of oscillator stability 
were covered in chapter 3. A number of 
sample circuits were presented, many of 
them offering excellent long-term sta- 
bility for use in transmitter applications. 
For die simpler receivers, these oscil- 
lators are generally adequate. 

Problems appear in the design of 
wide-dynamic-range receivers which 
make the general criteria in chapter 3 
(for obtaining stability) less dian suf- 
ficient, and in some cases even incor- 
rect. The performance parameter we 
bypassed was that of oscillator noise. 

The phenomenon of noise in an 
oscillator output is best understood by 
considering how an oscillator would 
appear when viewed with an ideal spec- 
trum analyzer. The amateur may not be 
familiar with this instrument. A spec- 
trum analyzer is essentially a receiver 
which has been optimized for test pur- 
poses. Unlike die receivers used for 
communications, the output is a display 
* the face of a cathode-ray tube. The 



fij. 22 - Generalized diagram of an oscillator. 


instrument is swept, with the tuning 
knob used to set die frequency of 
interest at the center of the CRT screen. 
The spectrum analyzer is a calibrated 
instrument, with the vertical axis repre- 
senting the power delivered to the input 
at die frequency corresponding to the 
horizontal position of the display at 
diat instant. 

When we refer to a spectrum ana- 
lyzer as being ideal, we mean that it has 
an unlimited dynamic range and has no 
internally generated noise. Such in- 
struments do not exist. We will deal 
with these realities later. 

A generalized schematic of an oscil- 
lator is presented in Fig. 22. This circuit 
is die same as diat given in die earlier 
VFO discussion and is used to examine 
die criteria necessary for oscillation. 
Reviewing die Barkhausen criteria, we 
recall diat a signal at point A will be 
increased in level in the amplifier. Part 
of the output will be matched to die 
resonator by means of Zl. The signal 
across die resonator will be matched to 
die amplifier input by inclusion of Z2. 
A self -sustained oscillation will result if 
(1) the amplitude of die resulting signal 
at A is larger dian the original, and (2) 
die phase of the output signal from Z2 
is exactly the same as that initially 
impressed at point A. 

Now, how would this signal appear 
in our hypodietical ideal spectrum ana- 
lyzer? Our usual image of oscillator 
behavior suggests the analyzer output 
shown in Fig. 23. Here, there is no 
output at any frequency except that to 
which the oscillator is tuned. The shape 
of the response is merely the shape of 
the filter used in the analyzer. A more 
realistic picture is that shown in Fig. 24, 
which is much different than the one 
provided by the ideal oscillator. 

The first difference noted is that the 
broadband noise is higher in level. That 
is, the baseline of the display is not at 
die bottom of the screen, but is a few 
dB higher. The origin of this noise can 
be understood if we go back to the 
oscillator block diagram of Fig. 22. The 
network, Z2, will reflect some real 


resistive impedance to the input of die 
amplifier. A noise power of kTB is thus 
available at the input to the amplifier. 
The noise power at the output of the 
amplifier will just be kill multiplied by 
the amplifier noise factor and giin. (The 
details of these noise calculations were 
presented earlier in this chapter.) While 
this noise will cause problems in a 
receiver, it is necessary in order to begin 
oscillation when power is applied 
initially. 

The second difference between the 
two spectrum-analyzer representations 
is the “noise pedestal” surrounding the 
carrier in Fig. 24, which was not present 
in Fig. 23. This noise is usually at- 
tributed to phase variations in the sys- 
tem. The width of the noise pedestal is 
equal to the loaded 3-dB bandwidth of 
the resonator. When the noise breaks 
out of the broadband noise floor, it will 
increase at a rate of 6 dB per octave as it 
approaches the carrier of the oscillator. 

Consider an oscillator operating at 5 
MHz with a loaded resonator Q of 10. 
The noise pedestal will begin at 4.75 
MHz, and will drop back into the 
broadband noise floor at 5.25 MHz. The 
noise will be 6 dB above the noise floor 
at 4.875 MHz and 12 dB up at 4.938 
MHz. Eventually, the carrier of the 
oscillator appears within the passband 
of the analyzer and dominates the dis- 
play. 

If a very narrow bandwidth is used 
in the analyzer, with some oscillators, a 
point may be reached where the noise 
increases at a 9 dB per octave rale 
instead of the 6-dB figure. The addi- 
tional 3 dB is the result of l/Tnoise in 
the amplifier. 

It is interesting to study further the 
basic oscillator of Fig. 22. Assume that 
dc bias has just been applied to the 
amplifier. Immediately, noise will lesull 
at the output. It will be routed through 
the phase -shift networks and resonator 
where it is applied agiin to the input. 
Some filtering occurs in the resonator, 
so the noise spectrum is already con- 
fined somewhat. The amplified input 
noise is routed through the amplifier 
and resonator system repeatedly, always 
increasing in amplitude with each pass 
around the loop. 

If we were to extend this analysis, 
we would predict that the positive 
feedback in the oscillator would cause 
the level of the signal in the loop to be 
an ever-increasing function of time. This 
is, of course, impossible. Something 
must happen to cause the amplitude of 
the loop signal to stop and stabilize at 
some finite level. There are two 
mechanisms that will cause this to hap- 
pen: age or limiting. 

As an example of age, consider the 
FET oscillator of Fig. 25. As the voltage 
across the tank builds up, the voltage 
impressed on diode CR1 will increase. 


Advanced Receiver Concepts 125 










Fig. 23 - How a signal would appear on an 
"ideal" spectrum analyzer display. 



Fig. 24 - A more realistic example of that 
given in Fig. 23. 




Rectification will occur, causing a dc 
voltage to build up across capacitor Cl . 
This voltage is applied to the gate of the 
FET and will serve as bias. As the 
magnitude of this bias increases, the 
average gate voltage becomes more nega- 
tive, driving the FET toward pinchoff 
and thereby reducing the gain of the 
amplifier. Amplitude stabilization oc- 
curs when the net gain is just enough to 
sustain oscillation. 

Limiting, as a mechanism for ampli- 
tude stabilization, is demonstrated in 
the circuit of Fig. 26. This oscillator was 
designed for low-noise performance by 
L. Gumm, K7HFD, and operates at 10 
MHz. The voltage from the resonator, 
which is applied to tire base, causes the 
collector current to change. This 
changing collector current is coupled 
back into the resonator through a link 
which is arranged to yield the proper 
phase for positive feedback. The maxi- 
mum peak current that can be supplied 
to the link is the current standing in the 
transistor pair. This is defined by the 
emitter resistor and the inductor, which 
has the effect of making the current 
appear to originate from a constant 
current source. With the peak collector 
current well defined, the voltage across 
the tank is also well defined and limited. 

In general, limiting is preferred over 
age as an amplitude-stabilization mech- 
anism, especially in oscillators for criti- 
cal receiver applications. The reason is 
the same as the one which makes fm 
receivers immune to noise in the pres- 
ence of strong signals — amplitude 
variations, including a-m noise, disap- 
pear from the output. This is not the 
case with oscillators utilizing an internal 
age loop for stabilization (Fig. 25). 
When considering the broadband noise 
floor of an oscillator (Fig. 24), half of 
the noise is associated with random- 
phase variations, with the other half 
being attributed to amplitude variations. 
By the use of limiting, the amplitude 
noise is virtually eliminated, yielding a 
3-d B decrease in the noise floor. 

Additional comments about the 


K7I1FD circuit will illustrate other fea- 
tures of low-noise oscillators. The col- 
lector link consists of two turns, while 
the base is tapped only one turn up 
from the cold end. Hence, the signal 
voltage at the base is quite large - a few 
volts pk-pk. This is highly desirable. 

The reader will recall from our dis- 
cussion of noise in amplifiers, that the 
degradation in output signal-to-noise 
ratio resulting from internally generated 
noise decreases as tire input signal-to- 
noise ratio increases. The goal in an 
oscillator design is to maximize the 
output signal-to-noise ratio. Hence, a 
general rule of thumb emerges: The 
drive at the input to the amplifier 
should be as high as possible. In the case 


of bipolar- transistor oscillators, such as 
the K7HFD example, the only limit 
imposed is that the emitter-base break- 
down of the transistor should not be 
exceeded. Not only will this lead to a 
degradation of transistor beta in time, 
but will cause extreme amounts of noise 
to be generated from the Zener-diode 
action. 

The same argument with regard to 
emitter-base breakdown can be applied 
to buffer amplifiers following an oscil- 
lator. Class C operation is quite ac- 
ceptable and will preserve low-noise 
performance as long as emitter-base 
breakdown does not occur. 

It is important in the K7HFD oscil- 
lator that the resonator energy be re- 



Fig. 26 - Circuit of the K7HFD low-noise oscillator. LI is 1.2 mH and uses 17 turns of wire 
on a T68-6 toroid core. The tap is at 1 turn. Q at 10 MHz is 250. L2 is a 2-turn link over LI . 


126 Chapter 6 











slricted by current limiting in Ql. and 
not by voltage clipping. Should the 
transistor go into saturation, the tank 
would be loaded severely by the satura- 
tion resistance of Ql. and would in- 
crease the width of tire noise pedestal. 
In the configuration shown, the reso- 
nator has minimal external loading. This 
is due to the high output resistance 
presented by the collector. The loading 
at the base is also minimal, resulting 
from the extreme turns ratio used and 
the Class C operation of Ql. Class C 
operation implies that the base of Ql 
extracts energy from the resonator only 
during a small fraction of the oscillation 
cycle. 

The presence of saturation in oscil- 
lators using limiting is detected easily 
with simple equipment. If the transistor 
is going into saturation, the output 
power will change significantly as the 
operating voltage is varied. This does 
not occur with the K7HFD circuit. 

Measurement of Noise in 
Local Oscillators 

It would be straightforward to 
measure die level of noise from oscil- 
lators if die “ideal” spectrum analyzer 
were available. Unfortunately, such in- 
struments do not exist. The better 
spectrum analyzers have dynamic ranges 
of 80 to 100 dB and are priced well 
beyond the reach of an amateur. Any 
good oscillator will have a noise floor 
which is over 100 dB below the carrier 
in a communications bandwidth. Hence, 
if die sensitivity of the analyzer were 
increased to the point diat the noise 
could be seen, the analyzer would be 
overloaded. The answer to the problem 
is to use an existing analyzer in con- 
junction with a crystal filter which has a 
center frequency near the oscillator 
output frequency. 

Shown in Fig. 27 is the system used 
for evaluation of the K7HFD oscillator. 
A 10-MHz filter with a 3-kHz band- 
width (6 poles) was used in conjunction 
with a Tektronix 71.12 Spectrum Ana- 
lyzer and a frequency counter. The 
crystal filter had a skirt response which 
caused the attenuation 10 kHz away 
from the center to be over 80 dB. The 
counter was used to set the oscillator to 
10.010 MHz and the output at 10.000 
MHz was observed in the analyzer. 
Because of the attenuation of the filter, 
the carrier of the LO was not over- 
loading the analyzer and the noise could 
be measured. The result was that the 
noise was over 120 dB below the output 
of 50 mW (+17 dBm) in a 3-kHz 
bandwidth, 10 kHz away from the 
carrier. 

The results of LO noise can be 
observed readily in some receivers. This 
results from the multiplier nature of 
mixers. That is, a mixer is a device with 
^ an output voltage which is proportional 


to the product of the two input signals. 
If a receiver with a very steep-skirted 
filter is tuned to a strong carrier, a 
clean -sounding tone is usually heard. 
However, as the receiver is tuned 
slowly away from the carrier, a point 
will be reached where there is no longer 
a clean tone coming from the receiver. 
Here, the attenuation of the crystal 
filter has suppressed the carrier signal. A 
noise output is. sometimes, still present. 
This will be the result of the strong 
carrier at the mixer rf port, mixing with 
the noise from the LO. 

It should be emphasized that the 
foregoing observation is based upon the 
assumption that the input strong carrier 
applied to the receiver is virtually noise- 
free. In a laboratory experiment this 
cleanliness is obtained by using a high- 
quality signal generator in conjunction 
with a narrow bandwidth (50 Hz or less) 
multipole crystal filter. This will ensure 
that the observed noise is a result of the 
local oscillator and not the noise output 
of the signal generator. 

On-the-air listening experiments can 



Fig. 27 - Sysiem tor evaluating the oscillator 
of Fig. 26. 


be enlightening. In one series of tests at 
W7Z01, a receiver using an FET oscilla- 
tor was used. With a 4-pole, 500-Hz- 
widc crystal filter as the main selectivity 
element, the receiver sounded excep- 
tionally clean. However, when a 10-pole 
filter with the same bandwidth was 
substituted, the effects of noise modula- 
tion were observed readily. 

Just as signal-generator noise was 
critical in a laboratory evaluation, the 
character of strong received signals is 
observable. As the receiver becomes 
more sophisticated, it is possible to 
detect subtleties in signal quality that 
would not be noticed in a more mun- 
dane receiver. 

There is one final experiment that 
can detect the presence of phase or 
frequency modulation in a receiver LO. 
This involves the use of a triggered 
audio-frequency oscilloscope, an instru- 
ment found in some amateur shops. A 
clean signal, such as might come from a 
crystal oscillator, is tuned with the 
receiver, and the audio output is moni- 
tored with the oscilloscope. The left 
side of the trace will always be clean 
that's the point where the sweep in the 
'scope is triggered. However, if fin noise 


is present in the receiver LO. the right- 
hand end of the trace will appear fuzzy. 
The time base of the oscilloscope should 
be set to display several cycles of audio. 
(Audio discriminators could be used for 
more exacting measurements.) 

General Design Criteria 

Using the above analysis it is possible 
to formulate a number of general rules 
for the design of quiet oscillators for 
critical receiver applications. 

1 ) Use as high a loaded resonator Q 
as can be obtained. This means not only 
that the unloaded Q should be high, but 
that the external loading by the oscilla- 
tor be minimal. Also, the high Q„ 
requirement often dictates the use of 
toroids which might have compromised 
temperature properties. 

2) Drive the input to the amplifying 
device as hard as possible without ex- 
ceeding any breakdown specifications. 
This also implies that the resonator 
should operate with high amounts of 
stored energy and the attendant large 
circulating currents. The high currents 
along with the first criterion will prob- 
ably compromise the long-term stabil- 
ity, making temperature compensation 
necessary. 

3) The transistor or FET should 
have capabilities to operate at fre- 
quencies very much higher than the 
operating frequency. This ensures not 
only that the device will have adequate 
gain, but will exhibit minimum unde- 
sired phase shift. This keeps the phase 
shift in the resonator and impedance 
matching networks (Fig. 22A) where 
they belong. For the same reason, single 
transistor or FET oscillators are pre- 
ferred over those using a multiplicity of 
devices. (This does not preclude buffer 
amplifiers.) 

4) While good output buffering is 
desirable, it is not generally necessary 
for receiver applications that the output 
be a pure sine wave as was advocated for 
transmitter VFOs. The reason for this is 
that most good mixers that is mixers 
with low 1MD — will create harmonics 
anyway. The undesired effects of these 
harmonics must be eliminated with 
proper choice of receiver i-f amplifier 
frequency and proper front-end prese- 
lection. With some diode mixers a 
square-wave LO is desired for least 
distortion. The LO waveform should be 
symmetrical, however, since an asymme- 
try can destroy the balance of an 
otherwise well-balanced mixer. 

Practical Examples 

There are a number of oscillators 
which will fulfill the foregoing criteria. 
How well they need to perlonn will 
depend upon the nature of the receiver 
being designed. Many of the simpler 
receivers in this book use straight- 
forward LOs. In no case has the receiver 

Advanced Receiver Concepts 127 








Fig. 28 — An oscillator which employs an 
MC1648P 1C. LI is a ferrite bead with two 
turns of wire. L2 is a high O u toroid tuned 
circuit. The link should contain only the 
number of turns necessary to sustain oscil- 
lation. A typical turns ratio is 4 : 1 , 


dynamic range been compromised by 
the oscillator. 

In several cases shown in the text. 
FET oscillators have been used. When 
operated at low frequencies (in the 2- to 
3-MHz range), they are quite suitable. If 
moved to higher frequencies, where it 
becomes harder to obtain a small loaded 
bandwidth, they may not be as appro- 
priate. 

Many of the FET oscillators in this 
book use the Clapp circuit in place of 
the simpler Colpitts one. This is desir- 
able from a noise standpoint. A detailed 
analysis of the Clapp network shows 
that the stored energy in the resonator 
is much larger than with the usual 
Colpitts design. 

The K7HFD oscillator used in the 
preceding discussion (Fig. 26) is one of 
the best that we have investigated. This 
oscillator operates with from 50 to 65 
volts pk-pk across the resonator, so 
some temperature compensation will 
probably be needed. This can be done 
with N750 ceramic capacitors as part of 
the tank capacitance. Also, several fer- 
rite beads were used in the original 
design in order to suppress vhf parasitic 
oscillations. 

Shown in Fig. 28 is an oscillator 
using a Motorola MC1648P integrated 
circuit. This chip was designed specifi- 
cally for oscillator applications and 
offers fair performance. In order to 
obtain the lowest noise output from this 
device, it is necessary that link coupling 
to the tank be employed. This is be- 
cause the internal circuitry of the IC is 
such that the maximum pk-pk voltage 
that can be obtained across the tank 
terminals is about 1.4. This would se- 
verely limit the stored energy in the 
tank. 

Link coupling between the tank and 

128 Chapter 6 


the MCI648P presents a problem that 
can make the chip difficult to use. This 
is the very high-frequency capability of 
the device. Because of this, the circuit is 
very prone to oscillate at a frequency 
determined by the inductance of the 
link and the stray capacitances. These 
vhf oscillations are usually killed with 
judicious use of a ferrite bead in series 
with the link. Often two turns of wire 
through the bead are required. A short 
lead length is mandatory, also. 

The MCI 648 P has a built-in age 
loop. For best spectral purity this 
should be defeated, and is accomplished 
by connecting a 1000-ohm resistor be- 
tween the +5-volt supply and pin 5. If a 
sine-wave output is desired, a resistor 
connected between this pin and ground 
can be used. Experimentation will be 
required to determine the proper value. 

If pin 5 is shorted to ground, oscilla- 
tion will cease. This characteristic can 
be useful in a multiband design where 
several oscillators might be used, one for 
each band (see Fig. 29). All of the 
outputs may be connected directly to- 
gether. Then, all of the oscillators ex- 
cept the one being used may be inhib- 
ited. This is easily done with a 
saturated-transistor switch. 

The output of the MC1648P is only 
a little more than I mW, which is too 
low for most diode mixers. The output 
may be increased through the use of a 
broadband amplifier. This approach is 
used in a transceiving system described 
later in the book. Alternatively, the 
output stage of the IC may be operated 
at a higher supply potential. The reader 
should consult the Motorola literature 
for this application. 

Since the MC1648P is capable of 
operation well into the vhf spectrum, 
careful bypassing and grounding tech- 
niques should be used. If high quality 
0.1 -^F capacitors are not available, the 
builder should use a .001 -pF capacitor 
in parallel with the larger value shown in 
the figure. Double-sided pc board is 
recommended. 

In any of the receiver LOs discussed, 
good power-supply regulation is needed. 
It is highly preferred if a separate 
voltage regulator be used on the pc 
board containing the oscillator. Special 
attention should be devoted to the 
rejection of power-supply hum. A high- 
gain active voltage regulator circuit is 
preferred over a simple Zener diode. If a 
Zener diode is used, it should be by- 
passed with a large electrolytic capaci- 
tor. 

Ideally, a receiver local oscillator 
should be well shielded in an rf-tight 
box. It does little good to carefully 
preselect and shield a receiver front-end, 
only to end up with spurious responses 
resulting from vhf signals finding their 
way to the mixer along the LO line. 

All of the arguments outlined here 


apply equally to BFOs used to drive a 
product detector. Good noise character- 
istics can be achieved easily with a BFO 
by using crystal control. The high un- 
loaded Q of a crystal eases the design 
considerably. Generally, any of the 
crystal-oscillator circuits described in 
chapter 2 are suitable, although shield- 
ing and decoupling requirements still 
apply. Examples of tunable BFOs are 
given in several of the construction 
projects in the book. 

Crystal-Controlled Converters 

Often it is desired to extend the 
tuning range of a receiver to bands other 
than those covered by an existing re- 
ceiver. This is done easily by the addi- 
tion of a crystal-controlled converter 
ahead of the receiver. All of the basic 
concepts outlined in previous examples 
will be presented and some philosophy 
will be added on our approach to the 
design of high-performance vhf conver- 
ters. 

Shown in Fig. 30 is the block dia- 
gram of a typical converter. A preselec- 
tor network is used at the input, tuned 
to the band of interest. The output of 
this is applied to an rf amplifier and 
then to another filter. The second filter 
is important in order to keep noise at 
the image frequency from reaching the 
mixer. Consequently, this filter is often 
called an “image-stripping filter.” The 
resulting signal is applied to a mixer. 
The mixer is driven by a crystal- 
controlled oscillator in order to provide 
stability and frequency accuracy. If a 
tuned output is used for the mixer, a 
multipole bandpass filter is a good 
choice if a wide tuning range is to be 
covered. 

In many situations the rf amplifier is 
not needed. This will depend upon the 
noise figure desired. Furthermore, in 
some converters it is desirable to dis- 
pense with the rf amplifier, but to 
include a post-mixer amplifier. This is 
done to preserve dynamic range of the 
overall system. 

Shown in Fig. 31 is the schematic of 
a simple converter for the 160-meter 
band. At 1.8 MHz the noise levels are 




Fig. 29 - Method for easy band switching ot 
MC1648P oscillators. 






Fig. 30 - Block diagram of a typical crystal-controlled converter. 


extremely high. As a result, it is point- 
less to strive for a low noise figure. 
Because of this, no rf amplifier is used, 
and the preselector is adjusted for a 
loaded Q of near 200. The output of the 
dual-gate MOSFET mixer is at 7 MHz. 
Although simple, this converter has per- 
formed well on “top-band.” No spuri- 
ous responses from broadcast stations 
have been detected, and the dynamic 
range has been adequate for some con- 
test operations. All continents except 
Africa have been received with this unit 
from Oregon, indicating adequate sensi- 
tivity. 

Shown in Fig. 32 is a simple conver- 
ter for the 6-meter band. In this case, a 
diode-ring mixer is preceded by a two- 
pole bandpass filter. The preselector 
was adjusted for a bandwidth of I MHz 
and had an insertion loss of 1 dB. The 
output of the diode ring is applied to a 
low-noise 14-MHz amplifier (see Fig. 
19), and then to the receiver used as the 
tunable i-f. The oscillator operates with 
a 36-MHz third overtone crystal and 
delivers +13 dBm to the diode ring. 
Careful measurements have not been 
performed on this converter. However, 
the noise figure appears to be about 10 
dB. The sensitivity is adequate to hear 
background noise when using a 2- 
element Yagi antenna. Of major signifi- 
cance is that there are no spurious 
outputs from channel 2 TV, even 
though the converter is used in a strong 
signal area. The usual level of channel 2 
on the 2-element Yagi is 0 dBm. One 
spurious response resulted from a local 
fm broadcast station. Its signal was 
converted to the 14-MHz band as a 
result of third-harmonic conversion in 
the diode-ring mixer. This response was 
eliminated by adding a low-pass filter. 

A similar approach to converter de- 
sign is presented in a later example. This 
family of converters is used to extend 
the coverage of a high-performance 
160-meter receiver to the high- 
frequency bands. 

VHF Converters 

A popular application of the crystal- 

J controlled converter is for reception in 
the vhf and uhf bands. Most converters 


used one to three stages of rf amplifica- 
tion, an active mixer, and often a 
post-mixer amplifier. The local- 
oscillator injection voltage was devel- 
oped with a low-frequency crystal and a 
frequency-multiplier chain. Usually, the 
circuitry was contained on an open 
chassis. 

While converters of that type were 
satisfactory once, times have changed. 
The vhf spectrum has become more 
heavily used. As a result, dynamic-range 
considerations are more important to- 
day than before. Furthermore, current 
interest in the reception of very weak 
signals, such as those encountered in 
moonbounce communications, places a 
severe constraint on noise figures. The 
following guidelines are offered for the 
design of high-performance vhf conver- 
ters. While each point will not be 
justified, the reader will see that they 
are all consistent with the design infor- 
mation presented for hf receivers. 

1 ) Use the highest frequency crystal 
in the LO that can be purchased. For 
example, if a 2-meter converter is built 


for 28-MHz output, use a 116-MHz 
crystal. If a frequency-multiplier chain 
is necessary (for example, a 432 conver- 
ter), use balanced multipliers and exten- 
sive output filtering. All subharmonic 
components should be attenuated at 
least 60 dB. 

2) Use diode mixers. Make sure that 
they are performing as desired. Provide 
diplexers at the i-f port to ensure image 
termination. 

3) Use low-noise methods at the i-f 
to provide a reasonably low system- 
noise figure at the mixer input. 

4) All rf amplifiers should be in 
separate, well-shielded containers with 
coax cables for interconnection. This 
will allow each stage to be matched and 
optimized individually. Use broadband 
techniques so that system stability is 
maintained at all frequencies. 

5) Use as much preselection as poss- 
ible. The input filter should have at least 
two poles, and the insertion loss should 
not exceed 1 dB. The image-stripping 
filter can have higher loss, but should 
have good stopband rejection. Helical 
resonators are recommended for the 
2-meter band, while interdigital filters 
are suggested for frequencies above 432 
MHz. 

6) Use extensive interstage shielding 
and decoupling of power supplies. Each 
stage should be packaged in its own 
container. High-quality feedthrough 
capacitors should be used for power 
supply connections. 

The techniques outlined are typical 
of those used in the communications 
industry. This is especially true for the 
construction of receivers for deep-space 
work, or for high-performance vhf and 
microwave instrumentation. Many of 



Fig. 31 - Circuit tor a simple 160-meter converter. LI has 33 turns of No. 22 wire, center 
tapped, on a T1 06-2 toroid. L2 is a 1 -turn link. L3 has 30 turns of wire on a T50-2 core. L4 
is a 5-turn link. L5 has 40 turns of wire on a T50-2 toroid, and L6 is a 7-turn link. Qt can be 
an MR FI 30 or a 40673. 


Advanced Receiver Concepts 129 









twt ; 

rh 7 /-r-7 I /—r~7 


BPF, IL -IdB. BW • IMHi 
S M . SILVER MICA 



0T vX 




i 


14 MHz 
OUTPUT 




+13 dBm LO 




SIMPLE SIX-METER CONVERTER 


Fig. 32 — Circuit of a simple 6-meter converter. LI and L2 have 8 turns of No. 18 wire, have an ID of 3/8 inch, and are 1 inch long. Tap at 1 
turn. Cl is 0.3 inch of FIG-174 coaxial line (C = approx. 0.5 pF). 


the suggestions can be ignored for casual 
applications. However, spurious re- 
sponses may result. 


Digital Frequency Readout 

A problem that has plagued the 
receiver builder was the construction of 
a frequency-readout mechanism. Not 
only were accurate and attractive dials 
difficult to build in the home shop, but 
they often caused the circuit design to 
be compromised. For example, some 
builders elected to build a dual- 
conversion receiver instead of a single- 
conversion one they regarded the 
virtues of a linear tuning scale with good 
resolution and accuracy to be worth the 
resultant degraded dynamic range. Such 
a compromise is no longer necessary. 

A modern approach to frequency 
readout is the use of digital circuitry 
with electronic display. Additional cir- 
cuits are required. However, mechanical 
construction problems are avoided. With 
a digital readout there is no need to 
couple a dial to the main tuning capaci- 
tor. Linearity of tuning is of little 
consequence. Long-term stability re- 
quirements may even be relaxed. While 
a moderate amount of circuitry is need- 
ed to realize a digital readout, the design 


is straightforward and construction is 
elementary. 

The virtues of digital readout do not 
come without a penalty. High-speed 
digital logic can create a large amount of 
rf noise. Some of this noise is broad- 
band in nature, while some is related to 
the discrete clock frequencies used in 
counters. Special precautions must be 
taken to keep this noise from creating 
spurious responses within the receiver. 

We will not attempt to cover in 
depth the theory of digital-logic design. 
There have been innumerable articles 
and books published on the subject (see 
bibliography). In this section we will 
confine our discussion to those details 
which are applicable to receivers. The 
barest fundamentals will be reviewed. A 
receiver using digital readout is pre- 
sented later in the book. 


Frequency-Counter Fundamentals 

Shown in Fig. 33 is a block diagram 
of a fundamental frequency counter. It 
consists of two sections: the signal 
counter and a time base. 

A time base consists of a crystal- 
controlled oscillator (often at 1 MHz) 
and a frequency divider. The circuit of 
Fig. 33 employs a division ratio of 


1000. This produces an output of 1 
kHz. The output of this divider is 
divided again by 2, yielding a string of 
pulses which are 1 ms wide. This signal 
occurs at point A in the figure. 

The 1 -ms-wide pulse is applied to an 
AND gate. The other input to the gate is 
the signal to be counted. Assume that 
the incoming frequency to be counted 
was 1.2 MHz. In a 1-ms period this 
signal will undergo 1200 complete tran- 
sistions. If the counters that follow the 
gate are set to 0 prior to application of 
the output of the gate, they will count 
up to 1200 during the 1-ms “timing 
window.” One decade counter is 
labelled LSD, standing for least signifi- 
cant digit. The last counter in the string 
is the most significant digit (MSD). 

The outputs of the decade counters 
are in a binary-coded decimal (BCD) 
format. There are four lines which can 
each take on a digital 0 to 1 . The BCD 
outputs are applied to elements termed 
"latches.” These are memory elements. 
Each IC package is actually a quad latch 
with one memory element for each BCD 
line from the counters. When a “strobe" 
line on the latches is activated with a 
positive voltage, the logic state present 
at the latch input is connected to the 



at 
time 
read o 


130 Chapter 6 


CO O 





lIT 



d <j 

s BESET 


dt dt 

^ TO COUNTERS 


1 > TO LATCHES 



L.S.0. 

1 1 

M.S.D. 



Fig. 33 — Block diagram of a fundamental frequency counter. 


output. When the strobe input again 
goes low, the information in the latch at 
that instant is retained. A signal to 
strobe the latches is derived from the 
1-ms time-base pulse. The trailing edge 
of the gate timing window is differenti- 
ated. This leads to a short pulse that 
follows the gate-control pulse. 

The output of the strobe pulse is 
also differentiated. This leads to another 
short pulse which follows the strobe 
action. This pulse is applied to the 
counters to reset them to 0, making 
them ready for the next burst of input 
data. 

The latches "remember” the state of 
the counters at the end of the counting 
period. The latch outputs are applied to 
ICs called decoder/drivers. They serve a 
dual function. First, they convert the 
BCD information supplied from the 
latches to the appropriate format to 
drive 7-segment light-emitting diode 
(LED) displays. Second, they provide 
enough output power to drive the LED 
displays. 

In the example described in Fig. 33, 
four digits were displayed, and a 1-ms 
timing window was used. The display 
was updated once in each 2-ms period. 
If the 1.2-MHz input was measured, the 
display would read “1200.” The read- 
out is in kHz. 

If the timing window was extended 
to I second, the results would be quite 
different. (This is realized by adding 
another divide-by-1000 chain to the 
time base.) The counter would then 
read out in Hz. The display would read 


"0000.” What has occurred is that the 
MSD counter has changed state 1000 
times during the period, ending up at 0. 
If the input frequency departed from 
1.2 MHz by, say, 2 Hz positive, the 
output would read “0002.” 

Assume that the time base is 1 ms, as 
shown, and that the input frequency is 
increased to 16.15 MHz. In this instance 
the output would read 6150. The lead- 
ing 1 , signifying the 1 0-MHz part, would 
have overrun the counter. This in no 
way decreases its utility. If it were 
desirable to read out the 10 MHz and 
higher frequencies, an additional 
counter, latch, decoder and LED could 
be added. Alternatively, the time base 
could be changed to 1 00 microseconds. 

One major problem occurs with the 
counter shown in Fig. 33. The display is 
updated once each 2 milliseconds. The 
human eye can only respond to changes 
that occur within about 100 ms. Be- 
cause of this, the display will appear to 
flicker in the LSD position. This will 
occur even if the stability of ail signals 
was uncompromised in stability, so long 
as they were not coherent. Additional 
circuitry will allow the display update 
period to be extended. 

Receiver Applications 

The counter just described is suitable 
for general-purpose applications. How- 
ever, it is not sufficient for receiver use. 
There are a number of reasons for this. 
The main one is that the frequency to 
be counted is not the incoming fre- 
quency but that of the local oscillator. 


The two frequencies differ by the i-f. 
Sometimes this is of no consequence. 
For example, if the i-f is exactly at a 
frequency that is divisible by 1 MHz, 
the LO can be counted directly. The 
digits that represent the MHz part are 
not displayed. This is especially effec- 
tive for a cw receiver. 

Even if the i-f lies at an exact 
multiple of 1 MHz, difficulties arise 
where ssb receivers are concerned. This 
is because the frequency of interest in 
ssb is not that at the center of the 
information being transmitted, but that 
of the suppressed carrier. This corre- 
sponds to the sum or the difference of 
the receiver LO and BFO. In principle 
these two oscillators could be mixed 
appropriately, and the resultant infor- 
mation counted. This method can work 
well if excessive shielding is used, which 
is possible. If the shielding and isolation 
are not nearly perfect, the receiver will 
respond to the mixed product which is 
precisely at the frequency being re- 
ceived. 

A cleaner approach is through the 
application of additional gates. Assume, 
for example, that the frequency to be 
counted corresponded to the sum of the 
BFO and the LO. A suitable display 
could be achieved by first counting one 
oscillator and then the other. The coun- 
ters would not be reset after switching 
between the first and the second. The 
result would correspond to the sum of 
the frequencies. 

If the desired output was the differ- 
ence of the BFO and the LO, additional 


Advanced Receiver Concepts 131 








difficulty would be enountered. This 
may be circumvented by use of up- 
down counters as well as with appro- 
priate gating. As pulses arrive at the 
input to a normal decade counter, the 
output follows the following sequence: 
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 0 and so 
forth. That is the unit counts up, 
starting at 0. Some more-elegant ICs 
have two inputs. One is the count-up 
input described. The other is a count- 
down input. Starting at 0, arriving 
pulses would cause it to read sequen- 
tially: 0, 9, 8, 7 and so forth. By using 
each of the inputs in a properly con- 
trolled way, one can obtain a result that 
corresponds to the difference of two 
frequencies. Multi-conversion systems 
may also be accommodated with these 
methods. 

Another method that may be used 
to read frequency more accurately is by 
use of presettable counters. In the fun- 
damental system of Fig. 33, the coun- 
ters were reset to 0 at the end of each 
counting period, after the information 
had been strobed into the latches. Pre- 
settable counters are more flexible. With 
the application of the proper pro- 
gramming signals, the “reset” pulse will 
set them to any desired output. Count- 
ing then commences from that point. 
By choosing the proper preset input, the 
offsets resulting from the i-f may be 
accommodated. 

The use of presettable counters is 
generally more direct. However, it is 
subject to any errors that may occur in 
the BFO frequency. The up-down coun- 
ter method automatically accommo- 
dates these drift and aging effects. 

Counter Noise Considerations 

If a frequency counter is to be used 
with a receiver, there are several pre- 
cautions that must be taken. If they are 
not, the noise from the counter may 
dominate the receiver output. Some of 
the problems are outlined below. 

The interface between the oscillators 
being counted and the digital circuits 
should be exceptionally clean. FKT buf- 
fers are suggested. The oscillators may 
be attenuated significantly and then 
reamplified to further enhance the isola- 
tion. 

Extensive shielding should be used. 
Ideally, the counter circuitry should be 
in an rf-tight box. High quality feed- 
through capacitors should be used for 
power supply lines. The 5-volt power 
supply often used for the digital circuits 
should be decoupled well from the 
receiver power supply. Often, some of 
the shielding recommendations may be 
relaxed if the rest of the receiver is well 
shielded. This would be required for 
other reasons in a high-performance 
receiver. 

Multiplexed displays should be 
avoided. This requires some explana- 


tion. The decoder/drivers used in the 
circuit of Fig. 33 all operate in parallel. 
The signals sent to the LED displays are 
dc ones that change only when the 
display is updated. In contrast, there are 
many displays and matching decoder/ 
drivers that operate in a sequential 
manner. This allows the outputs of 
several sets of latches to be applied to a 
single decoder/driver at one time. Simi- 
larly, a fewer number of output lines 
are required to attach to a collection of 
LED segments. The various digits are 
scanned at a high rate and pulsed on for 
short periods. The eye perceives all of 
the digits as being on, simultaneously. 
Most digital clocks and pocket cal- 
culators use multiplexed displays. The 
fact that high-speed digital circuits are 
changing state continually leads to large 
noise outputs. 

The crystal oscillator used as the 
clock for the time base should not be 
related directly to the receiver i-f. For 
example, a receiver built by one of the 
writers uses a 9-MHz i-f and a digital 
readout. When the counter was first 
constructed, a 1-MHz clock was used. 
The ninth harmonic could be heard 
faintly in the i-f (at a very low level 
corresponding to an input signal of 
— 138 dBm). The clock was moved to 2 
MHz, thereby solving the problem. The 
seventh harmonic can be heard at 14 
MHz only when an antenna is connected 
to the receiver. 

A final precaution is to time- 
sequence the time base. This is realized 
in the counter of Fig. 33 by placing a 
gate between the crystal oscillator and 
the divide-by-1000 counter. The oscilla- 
tor is allowed to run continuously. 
However, the divider circuit is on only 
when it is needed. If the display up-date 
rate is made slow (1/2 second), there is 
no digital circuitry operating during 
most of the listening time. In the 
extreme, provision could be made to 
completely shut the counter circuits off 
by means of a front-panel switch. 

High-Resolution Frequency Readout 

The use of a counter as the fre- 
quency display in a receiver has a 
number of advantages. Many have been 
outlined. One is the high resolution of 
the counter, which allows the receiver 
to be reset precisely to a previously 
logged frequency. The limit is the inter- 
val used for the time-base and the 
short-term stability of the oscillators. 

While reset ability may be high, 
similar accuracy in readout is not im- 
plicit. First, there may be some drift in 
the clock oscillator used in the time 
base. Of greater significance is the band- 
width of the receiver. For example, if a 
500-Hz-bandwidth receiver is used with 
a digital readout, the accuracy of a 
received signal is, at best, 500 Hz. The 
receiver may be tuned over a 500-Hz 


range, leading to a 500-Hz change in the 
readout while still copying an arriving 
signal. 

There is a method that may be 
employed to extend the accuracy of a 
digital readout. Auxiliary equipment is 
required, which is constructed easily or 
integrated into an existing receiver. 
Assume that the receiver counter has a 
time base with a 1 -second counting win- 
dow. The resulting resolution is 1 Hz. 

The first extra piece of equipment is 
a 1-MHz standard. This unit is set 
carefully against WWV or some other 
standard of known accuracy. After the 
transfer standard is calibrated, an har- 
monic is tuned with the receiver. Once 
in the passband, the receiver is tuned 
until the readout displays an exact 
multiple of the 1-MHz standard. For 
example, on the 20-meter band, the 
readout would read 14.000000 MHz. 
With the receiver so tuned, an external 
audio oscillator is adjusted to produce 
exactly the same audio frequency. The 
comparison may be done with an oscil- 
loscope (in the X-Y mode using Lis- 
sajous patterns), with a digital phase- 
frequency detector, with the counter, or 
even by ear. 

Once the pitch calibration is per- 
formed, an unknown signal may be 
tuned to produce exactly the same pitch. 
When this is realized, the precise fre- 
quency is read directly. On several 
occasions one of the writers achieved 
1-Hz accuracy in W1AW Frequency 
Measuring Tests with this technique. 

It should be mentioned that this 
method appears to be more accurate 
than those using a “zero-beat” compari- 
son. Also, the receiver used for these 
tests had sufficient i-f selectivity that 
zero beat could not be detected. The 
ultimate limitation of this approach to 
frequency measurement is the short- 
term stability of the oscillators used and 
the inaccuracies related to Doppler shift 
during WWV calibration. 

A 1-Hz frequency accuracy is rarely 
needed for an amateur receiver. A ques- 
tion of more practical nature concerned 
the general usefulness of a digital read- 
out during routine communications. 
Would an analogue dial be missed? The 
writers’ answer to this query is an 
uncategorical no! The digital readout 
was found remarkably easy to adapt to. 
The ability to set the receiver on a 
known frequency for monitoring pur- 
poses has been immensely useful. 

A High-Performance Receiver 
for 160 Meters 

A high order of dynamic range is 
important to good reception in areas of 
high signal density. Operation on 160 
meters requires a better than average 
communicatibns receiver, particularly in 
situations where commercial a-m broad- 
cast stations are nearby, and when the 


132 Chapter 6 



Fig. 34 - Schematic diagram of the receiver front end. Fixed-value capacitors are disk ceramic 
position. All slug-tuned inductors are contained in individual shield cans which are grounded. 


Cl - Three-section variable, 100 pF per 
section. Model used here obtained as 
surplus. 

J1 - SO-239. 

J2 - Phono jack. 

L1.L4-38 to 68 pH, O u of 175 at 15 
MHz, slug-tuned (J. W. Miller 43A685CBI 
in Miller S-74 shield can). 

L2. L3 - 95 to 187 pH. Q u of 175 at 1 S 
MHz, slug tuned (J. W. Miller 43A154CBI 
in S-74 shield can). 

15, L6 — 1 ,45-pH toroid inductor, Q u of 
250 at 1.8 MHz. 15 turns No. 26 
enam. wire on Amidon T-50-2 
toroid. 


L7. L9 — 13-pH slug-tuned inductor (J. W. 
Miller 9052). 

L8 — 380-pH slug-tuned inductor (J. W. 
Miller 9057). 

L10 — 16 turns No. 30 enam. wire over L1 1 
winding. 

L1 1 - 45 turns No. 30 enam. wire on 
Amidon T-50-2 toroid, 8.5 pH. 

LI 2 — 42 -pH slug-tuned inductor, G„ of 50 
at 1.8 MHz (J. W. Miller 9054), 

L13 — 8.7-pH toroidal inductor. 12 turns 
No. 26 enam. wire on Amidon FT -37-61 
ferrite core. 

LI 4 — 120- to 280-pH, slug-tuned inductor 
(J.W. Miller 9056). 


unless otherwise noted. Resistors are 1/2-W com- 

L15— 1.3- to 3.0-mH, slug-tuned inductor 
(J.W. Miller 9059). 

Q1 , Q2, Q3 — Motorola JFET. 

RFC1 — 2.7-mH miniature choke (J. W. 

Miller 70F273AI). 

RFC2 — 10-mH miniature choke (J. W. 

Miller 70F102AI). 

SI — Three-pole, two-position phenolic 
wafer switch. 

S2, S3 — Two-pole, double-throw miniature 
toggle. 

U1 - Mini-Circuits Labs. SRA-1-1 doubly 
balanced diode mixer (2913 Quentin Rd., 
Brooklyn. NY 11229). 




operator lives near other 160-meter en- 
thusiasts who are active on the band. 
The effects of blocking, cross modula- 
tion, and IMD can render a poorly 
designed receiver useless in the fore- 
going situation, making weak-signal 
work an impossible task. 

Some ordinary design procedures 
can be followed when building a re- 
ceiver with above average dynamic-range 
parameters, and the construction job is 
not a difficult one. Special care must go 
into the front-end design and gain distri- 


bution of the receiver circuitry to assure 
the performance specified here, but 
construction of such a receiver should 
be no more exacting than would be the 
case when building a mediocre one. 

Although this is a single-band re- 
ceiver, coverage of 80 through 15 
meters can be accomplished with good 
dynamic-range traits by employing the 
converters described later in this chap- 
ter. They were designed for high perfor- 
mance also, and the desired chacteristics 
were based on the dynamic -range profile 


of this receiver. That is, the two systems 
are compatible by design intent. IMD of 
the main-frame receiver (tested at 1.9 
MHz) is -95 dB. Noise floor is -135 
dBm, and blocking of 1 dB occurs at 
some point in excess of 123 dB above 
the noise floor. With the mating 20- 
meter converter attached the IMD = 88 
dB, noise floor is -133 dBm, and 
blocking is in excess of 123 dB. The 
20-meter tests were performed with the 
fixed-tuned 160-meter front-end filter 
in the circuit. Tests for dynamic range 


Advanced Receiver Concepts 133 










Fig. 35 - Response curve of the tunable 
front-end filter, centered on 1 .9 MHz. 


on 160 meters were performed with the 
tunable Cohn filter in the circuit. This 
receiver was described first in QST for 
June and July, 1976. 

Front-End Circuit 

Fig. 34 shows the rf amplifier, 
mixer, and post-mixer amplifier. What 
may seem like excessive elaboration in 
design is a matter of personal whim, but 
the features are useful, nevertheless. For 
example, the two front-end attenuators 
aren't essential to good performance, 
but are useful in making accurate mea- 
surements (6, 12 of 18 dB) of signal 
levels during on-the-air experiments 
with other stations (antennas, ampli- 
fiers and such). Also, FL2, a fixed- 
tuned 1 .8- to 2 -MHz bandpass filter, 
need not be included if the operator is 
willing to repeak the three-pole tracking 
filter (FL1) when tuning about in the 
band. The fixed-tuned filter is con- 
venient when the down converters are in 
use. 

The benefits obtained from a highly 
selective tunable filter like FL1 are seen 
when strong signals are in or near the 
160-meter band. The rejection charac- 
teristics can be seen in Fig. 35. Insertion 
loss was set at 5 dB in order to narrow 
the filter response. In this example the 
high-Q slug-tuned inductors are isolated 
in aluminum shields, and the three- 
section variable capacitor which tunes 
them is enclosed in a shield made from 
pc-board sections. Bottom coupling is 
accomplished with small toroidal coils. 

Rf amplifier Q1 was added to com- 
pensate for the filter loss. It is mis- 
matched intentionally by means of L10 
and Lll to restrict the gain to 6 dB 
maximum. Some additional mis- 
matching is seen at LI 2, and the mixer 
is overcoupled to the FET tuned output 
tank to broaden the response (1.8 to 2 
MHz). The design tradeoffs do not 
impair performance. 

Tire doubly balanced diode-ring 


mixer (Ul) was chosen for its excellent 
reputation in handling high signal levels, 
having superb port-to-port signal isola- 
tion, and because of its good 1MD 
performance. The module used in this 
design is a commercial one which con- 
tains two broadband transformers and 
four hot-carrier diodes with matched 
characteristics. The amateur can build 
his own mixer assembly in the interest 
of reduced expense. At the frequencies 
involved in this example, it should not 
be difficult to obtain performance equal 
to that of a commercial mixer. 

A diplexer is included at the mixer 
output (LI 3 and the related .002 capa- 
citors). The addition was worthwhile, as 
it provided an improvement in the noise 
floor and IMD characteristics of the 
receiver. The diplexer works in combin- 
ation with matching network L14, a 
low -pass L-type circuit. The diplexer is a 
high-pass network which permits the 
56-ohm terminating resistor to be seen 
by the mixer without degrading the 
455-kHz i-f. The low -pass portion of the 
diplexer helps reject all frequencies 
above 455 kHz so that the post-mixer 
amplifier receives only the desired infor- 
mation. The high-pass section of the 
diplexer starts rolling off at 1.2 MHz. A 
reactance of 66 ohms (X c and AT;.) was 
chosen to permit use of standard-value 
capacitors in the low-2 network. 

A pair of source-coupled JFETs is 
used in the post-mixer i-f preamplifier. 
The 10,000-ohm gate resistor of Q2 sets 
the transformation ratio of the L net- 
work at 200: 1 (50 ohms to 1 0 kf2). An 
L network is used to couple the pre- 


PF. 

C3 Miniature 30-pF air variable. 

CR1 — High-speed switching diode, silicon 
type 1N914A. 

L18 - 17- to 41 -pH slug-tuned inductor, 
Q„ of 175 (J. W. Miller 43A335CBI in 



The receiver is built in a homemade alumi- 
num cabinet. A two-tone gray and flat- 
black paint job has been applied. Black 
Dvmo tape labels are used for identifying 
the controls in the black area, and gray labels 
are affixed to the gray portion of the front 
panel. A cut-down Jackson Brothers vernier 
dial mechanism (two-speed) is used for fre- 
quency readout. 

amplifier to a diode-switched pair of 
Collins mechanical filters which have a 
characteristic impedance of 2000 ohms. 
The terminations are built into the 
filters. 

Gain distribution to the mixer is 
held to near unity in the interest of 
good IMD performance. The preampli- 
fier gain is approximately 25 dB. The 
choice was made to compensate for the 
high insertion loss of the mechanical 
filters - 10 dB. Without the high gain of 
Q2 and Q3 there would be a deteriora- 
tion in noise figure. 

Local Oscillator 

A low noise floor and good stability 
are essential traits of the local oscillator 


Miller S-74 shield can). 

LI 9 — 10- to 18.7-pH slug-tuned pc-board 
inductor (J, W. Miller 23A155RPC). 
RFC13, RFC14 - Miniature 1 -mH rf choke 
(J. W. Miller 70F103AI). 

VR2 - 8.6-V, 1 -W Zener diode. 



Fig. 36 — Circuit diagram of the local oscillator. Capacitorsare disk ceramic unless specified 
differently. Resistors are 1/2-W composition. Entire assembly is enclosed in a shield box 
made from pc-board sections. 

C2 - Double-bearing variable capacitor, 50 


134 Chapter 6 




in a quality receiver. The requirements 
are met by the circuit of Fig. 36. Within 
the capabilities of the ARRLIab measur- 
ing procedures, it was determined that 
VFO noise was at least 90 dB below 
fundamental output. Furthermore, 
stability at 25°C ambient temperature 
was such that no drift could be measured 
from a cold start to a period three 
hours later. Mechanical stability is excel- 
lent: Several sharp blows to the VFO 
shield box caused no discernible shift in 
a cw beat note while the 400-Hz i-f 
filter was actuated. VFO amplifier Q14 
is designed to provide the recommended 
+7 dBm mixer injection voltage. Fur- 
thermore, the output pi tank of Q 14 is 
of SO ohms characteristic impedance. 
Though not of special significance in 
this application, the measured harmonic 
output across 50 ohms is -36 dB at the 
second order, and -47 dB at the third 
order. 

Filter Module 

In the interest of minimizing leakage 
between the filter input to output ports 
(Fig. 37), diode switching is used. The 
advantage of this method is that only dc 
switching is required, thereby avoiding 
the occasion for unwanted rf coupling 
across the contacts and wafers of a 
mechanical switch. 1N914 diodes are 
used to select FL3 (400-Hz bandwidth) 
or FL4 (2.5-kHz bandwidth). Reverse 
bias is applied to the nonconducting 


diodes. This lessens the possibility of 
leakage through them. Because the 
Collins filters have a characteristic im- 
pedance of 2000 ohms, the output coup- 
ling capacitors from each are 120 pF 
rather than the low-reactance .0 1 -juF 
units, as used at the filter inputs. With- 
out the smaller value of capacitance the 
filters would see the low base imped- 
ance of Q4, the post-filter i-f amplifier. 
The result would be one of double 
termination in this case, leading to a loss 
in signal level. Additionally, the 120-pF 
capacitors help to divorce the input 
capacitance of the amplifier stage. The 
added capacitance would have to be 
subtracted from the 350- and 510-pF 
resonating capacitors at the output ends 
of the Filters. 

The apparent overall receiver gain is 
greatest during cw reception, owing to 
the selectivity of cw filter FL3. To keep 
the S-meter readings constant for a 
given signal level in the ssb and cw 
modes, R7 has been included in the 
filter/amplifier module. In the cw mode, 
R7 is adjusted to bias Q4 for an S-meter 
reading equal to that obtained in the ssb 
mode. Voltage for the biasing is ob- 
tained from the diode-switching line 
during cw reception. 

I-F Amplifier 

A receiver i-f system should be capa- 
ble of providing a specific gain, have an 
acceptable noise figure, and respond 



Considerable space remains beneath the chas- 
sis for the addition of accessory circuits or a 
set of down converters. At the upper left are 
the adjustment screws for the tunable filter, 
plus the bottom-coupling toroids. At the left 
center is the fixed-tuned front-end filter. To 
the right is the rf -amplifier module. A 100- 
kHz MFJ Enterprises calibrator is seen at the 
far lower left. Immediately to its right is the 
mixer/amplifier assembly. The large board at 
the lower center contains the i-f filters and 
post-filter amplifier. Most of the amplifier 
components have been tacked beneath the pc 
board because of design changes which oc- 
curred during development. 


satisfactorily to the applied age. This 
almost bromidic judgment is not as trite 
as it may seem, for some designers use a 
haphazard approach to this part of a 
receiving system. Two of the more 
serious shortcomings in some designs are 



Fig. 37 — Schematic diagram of the filter and i-f post-filter amplifier. Capacitors are disk ceramic. Resistors are 1/2-W composition. 

CR2-CR5, incl. — High-speed silicon switch- RFC3-RFC10, incl. — 10-mH miniature rf S4 — Double-pole, double-throw toggle or 

mg diode. 1N914A. choke (J. W. Miller 70F102A1 ). wafer. 

FL3 — Collins mechanical filter F455FD-04. R7 — Pc-board control, 10,000 ohms, linear T1 — Miniature 455-kHz i-f transformer 

FL4 - Collins mechanical filter F455FD-25. taper. |J. W. Miller 2067. 30.000 to 500 ohms). 


Advanced Receiver Concepts 135 










and af preamplifier are on a common circuit 
CR6-CR9, incl. — High-speed silicon, 

1 N914A or equiv. 

CR10 — Motorola MV-104 Varicap 
tuning diode. 

LI 6 — Nominal 640-pH slug-tuned 
inductor (J. W. Miller 9057). 

LI 7 — Nominal 60-pH slug-tuned 


oard, which is not shielded. 

inductor (J. W. Miller 9054). 

R1 — 100.000-ohm I inear -taper 

composition control (panel mount). 
RFC1 1 - 2.5-mH miniature choke (J. W. 
Miller 70F253A1 ). 

RFC12 — 10-mH miniature choke (J. W. 
Miller 70F102A1 ). 


T2. T3 — 455-kHz i-f transformer. See 
text. (J.W. Miller 2067). 

T4 - Trifilar broadband transformer. 15 
trifilar turns of No. 26 enam. wire on 
Amidon T-50-61 toroid core. 

U2, U3 - RCA 1C. 

VR1 - 9.1 -V. 1 -W Zener diode. 


poor age (clicky, pumping, or inade- 
quate range)'and insufficient i-f gain. 

A pair of RCA CA3028A ICs is used 
in the i-f strip. Somewhat greater gain 
and age range is possible with MC1590G 
ICs, and they are the choice of many 
builders. However, the CA3028As, con- 
figured as differential amplifiers, will 
provide approximately 70 dB of gain 
per pair when operated at 455 kHz. This 
gives an age characteristic from maxi- 
mum gain to full cutoff which is en- 
tirely acceptable for most amateur 
work. 

Fig. 38 shows the i-f amplifiers, 
roduct detector, and Varicap-tuned 
FO. Transformer coupling is used be- 
tween U2 and U3, and also between U3 
and the product detector. The 6800- 
ohm resistors used across the primaries 


of T2 and T3 were chosen to force an 
impedance transformation which the 
transformers can't by themselves pro- 
vide: Available Miller transformers with 
a 30,000-ohm primary to 500-ohm 
secondary characteristic are used. U2 
and U3 have 10- and 22-ohm series 
resistors in the signal lines. These were 
added to discourage vhf parasitic oscil- 
lations. 

Age is applied to pin 7 of each IC. 
Maximum gain occurs at +9 V, and 
minimum gain results when the age 
voltage drops to its low value, +2 V. The 
age is rf-derived, with i-f sampling for 
the age amplifier being done at pin 6 of 
U3 through a 100-pF blocking capac- 
itor. 

The 1000-ohm decoupling resistors 
in the 12-V feed to U2 and U3 drop the 


operating voltage to +9. This aids stabil- 
ity and reduces i-f system noise. The 
amplifier strip operates with uncondi- 
tional stability. 

Product Detector 

A quad of 1N914A diodes is used in 
the product detector. Hot-carrier diodes 
may be preferred by some, and they 
may lead to slightly better performance 
than the silicon units. A trifilar broad- 
band toroidal transformer, T4, couples 
the i-f amplifier to the detector at a 
50-ohm impedance level. BFO injection 
is supplied at 0.7 V rms. 

BFO Circuit 

In the interest of lowering the cost 
of this project, a Varicap (CRIOof Fig. 
38) is used to control the BFO fre- 


136 Chapter 6 




Fig. 39 — Schematic diagram of the age system, 
polarity is indicated, which signifies electrolytic. 
This module is not enclosed in a shield 
CR12.CR13 - High-speed silicon. 1N914A 
or equiv. 

010, Q1 1 , Q14 — Motorola transistor. 

R2, R4. R5 — Linear-taper composition pc- 
board mount control. 

R3 — 10.000-ohm linear-taper control, panel 


Capacitors are disk ceramic except when 
Fixed-value resistors are 1/2-W composition 

mounted. 

RFC15 — 2.5-mH miniature choke (J. W. 

Miller 70F253A1). 

S5 — Single-pole, single-throw toggle. 

U4 — Dual-in-line 8 pin 741 op amp. 

Ml — 0- to 1-mA meter. 


compartment. 


the Q10/Q11 gain is determined as: 
Gain (dB) = 20 log R c -r R s Control R2 
has been included as part of R, to 
permit adjustment of the age loop gain. 
Each operator may have a preference in 
this regard. The age is set so it is fully 
actuated at a signal-input level of 10 ,uV. 
Age action commences at 0.2 /aV (1 dB 
of gain compression). 

Age disabling is effected by remov- 
ing the operating voltage from QlOand 
Q1 1 by means of S5. Manual i-f gain 
control is made possible by adjusting R3 
of Fig. 39. Age delay is approximately I 
second. Longer or shorter delay periods 
can be established by altering the values 
of the QI4 gate resistor and capacitor. 
Age amplifier gain is variable from 6 to 
40 dB by adjusting R2. Age action is 
smooth, and there is no evidence of 
clicks on the attack during strong-signal 
periods. At no time has age “pumping" 
been observed. 

Audio System 

A major failing of many receivers is 
poor-quality audio. For the most part 
this malady is manifest as cross-over 
distortion in the af-output amplifier. 
Moreover, some receivers have marginal 
audio-power capability for normal room 
volume when a loudspeaker is used. 
Some transformerless single-chip audio 
ICs (0.25- to 2-W class) exhibit a prohib- 
itive distortion characteristic, and this is 


quency. Had a conventional system 
been utilized, three expensive crystals 
would have been needed to handle 
upper sideband, lower sideband, and cw. 
The voltage-variable capacitor tuning 
method shown in Fig. 38 is satisfactory 
if the operator is willing to change the 
operating frequency of the BFO when 
changing receive modes. Adjustment is 
done by means of front-panel control 
R1 . Maximum drift with this circuit was 
measured as 5 Hz from a cold start to a 
time three hours later. A Motorola 
MV-104 tuning diode is used at CR10. 

Q6 functions as a Class A BFO 
amplifier/buffer. It contains a pi- 
network output circuit and has a 50- 
ohm output characteristic. The main 
purpose of the amplifier stage is to 
increase the BFO injection power with- 
out loading down the oscillator. 

AGC Circuit 

Fig. 39 shows the age amplifier, 
rectifier, dc source follower, and op- 
amp difference amplifier. An FET is 
used at Q10 because it exhibits a high 
input impedance and will not, there- 
fore, load down the primary of T3 in 
Fig. 38. 01 is direct coupled to a pnp 
transistor, Qll. Assuming that R, and 
R2 are treated as a single resistance, R s , 



Top-chassis view of the receiver. The R-C active filter and audio preamplifier are built on the 
pc board at the upper left. To the right is the BFO module in a shield box. The age circuit is 
seen at the lower left, and to its right is the i-f strip in a shield enclosure. The large shield box 
at the upper center contains the VFO. To its right is the tunable front-end filter. The three- 
section variable capacitor is inside the rectangular shield box. The audio amplifier module is 
seen at the lower right. The small board (mounted vertically) at the left center contains the 
product detector. Homemade end brackets add mechanical stability between the panel and 
chassis and serve as a support for the receiver top cover. 


Advanced Receiver Concepts 137 





especially prominent at low signal levels. 
The unpleasant effect is one of “fuzzi- 
ness” when listening to low-level signals. 
Unfortunately, external access to the 
biasing circuit of such ICs is not typical, 
owing to the unitized construction of 
the chips. 

Since undistorted audio is an impor- 
tant feature of a quality communica- 
tions receiver, discrete devices have been 
employed in this circuit. The 
complementary-symmetry output tran- 
sistors and the op-amp driver are config- 
ured in a manner similar to that used by 
Jung in his Op Amp Cookbook. Maxi- 
mum output capability is 3.5 W into an 
8-ohm load. An LM-301A driver was 
chosen because of its low-noise profile. 
There has been no aural evidence of 
distortion at any signal level while using 
the circuit of Fig. 40. The rationale in 
this situation is one of having consider- 


ably more audio power available than is 
needed a practice used in hi-fi work. 

R-C Active CW Filters 

A worthwhile improvement in signal- 
to-noise ratio can be realized during 
weak-signal reception by employing an 
R-C active bandpass filter. A two-pole 
version (FL5) is shown in Fig. 40. A 
peak frequency of 800 Hz results from 
the R and C values given. 

The benefits of FL5 are similar to 
those described elsewhere in this vol- 
ume, where a second i-f filter (at the i-f 
strip output) is used to reduce wide- 
band noise from the system. The R-C 
active filter serves in a similar manner, 
but performs the signal “laundering” at 
audio rather than at rf. The technique 
has one limitation — monotony in 
listening to a fixed-frequency beat note, 
which is dictated by the center fre- 


quency of the audio filter. The R-C 
filter should be designed to have a peak 
frequency which matches the cw beat- 
note frequency preferred by the opera- 
tor. That is, if the BFO is adjusted to 
provide an 800-Hz cw note, the center 
frequency of FL5 should also be 800 
Hz. 

Experience with FL5 in this receiver 
has proved in many instances that weak 
DX signals on 160 meters could be 
elevated above the noise to a Q5 copy 
level, while without the filter solid copy 
was impossible. It should be stressed 
that high-£> capacitors be used from C4 
to C7, inclusive, to assure a sharp peak 
response. Polystyrene capacitors satisfy 
the requirement. To ensure a well- 
defined (minimum ripple) center fre- 
quency, the capacitors should be 
matched closely in value (5 percent or 
less). Resistors of 5-percent tolerance 



Fig. 40 - Diagram of the audio amplifier and R-C active filter. Capacitors are disk ceramic unless otherwise noted. Polarized capacitors are 
electrolytic or tantalum. Fixed-value resistors are 1/2-W composition. This circuit is not contained in a shield box. Heat sinks are used with 
Q8 and Q9. 

C4-C7, incl. — See text. J3 - Phone jack. S6 — Double-throw, double-pole toggle. 

CR11 - High-speed silicon. 1N914A or R6 — 10.000-ohm audio-taper composition U4 - National Semiconductor LM-301A 1C. 

equiv. control, panel mounted. U5 - Signetics N5558 dual op-amp 1C. 


138 Chapter 6 












Exterior view of the high-performance con- 
verter assembly. A gray and black spray-paint 
finish is applied to the homemade aluminum 
cabinet. Lettering is by means of a Dymo 
tape labeler. 


should be employed in the circuit, 
where indicated in Fig. 40. 

Summary Comments 

The photographs illustrate a modular 
construction technique. All rf-circuit 
assemblies are isolated from one 
another, and from outside energy influ- 
ences, by means of shield compart- 
ments. Signal points are joined (module 
to module) with RG-174/U subminia- 
ture coaxial cable, the shield braids 
being grounded to the chassis al each 
end. Feedthrough-type .001-qF capaci- 
tors are used at the 12-V entry points of 
the modules. The foregoing measures 
help to prevent birdies and unwanted 
stray rf pickup. 

The tuning range of the receiver is 
200 kHz. This means that for use with 
converters the builder will have to 
satisfy himself with the cw or ssb band 
segments. The alternatives are to in- 
crease the local oscillator tuning range 
to 500 kHz, or use a multiplicity of 
converters to cover the cw and ssb 
portions of each band. 

High-Performance Converters 

This section provides circuits for a 
group of converters (80 through 15 
meters) for use with the high- 
performance 160-meter receiver de- 
scribed in this chapter. These units were 
described originally in QST for June. 
1976. 

Converter Designs 

After a bit of number crunching it 
was concluded that the converters 
should have a net gain of about 10 dB 
and an output intercept of approxi- 
mately + 17 dBm or higher. For work on 
the bands up through 14 MHz, a noise 
figure of 13 to 16 dB was deemed 
acceptable. On the higher bands some 
compromise in dynamic range would be 
tolerable in order to achieve lower noise 
figures. In studying the available circuit 
combinations it was decided to 
base the front end of the converters on 
a diode-ring mixer. The mixer would be 



J, 


Fig. 41 — Block diagram of the CER-verters. 



Fig. 42 — Diagram of the mixer and amplifier. Fixed-value capacitors are disk ceramic unless 
noted otherwise. Resistors are 1f2-W composition. See tables for component values not marked. 
U1 is a ML-1 or SRA-1 doubly balanced diode-ring mixer assembly. L2 (1.62 xH) has 18 turns of 
No. 22 wire on a T50-2 toroid core. T1 primary has 50 turns of No. 22 wire on an FT-50-72 toroid i 
core. The secondary contains 7 turns of No. 22 wire. LI has 65 turns No. 26 enam. wire on a 
T68-2 toroid core. 



EXCEPT AS INDICATED, DECIMAL 
VALUES Of CAPACITANCE ARE 
IN MICROfARAOStjiF I ; OTHERS 
ARE IN PIC0EARA0S I »F OR j,jiF |; 
RESISTANCES ARE IN OHMS, 

> -IOOO.M.IOOO 000 



TO DIOOE-RING 
3 LO PORT 


Fig. 43 - Diagram of the filler and crystal oscillator used on 20, 40 and 80 meters. Numbered 
fixed-value capacitors are silver micas. Resistors are 1/2-W composition. See Tables 1 and 2 for 
parts values. 


Advanced Receiver Concepts 139 










Table 1 

BAND 

(MHz) 


3.5 to 3.7 


L3, L4. LB 
(TURNSCORE) 


L9 

(TURNSCORE) 


L5. L6, L7 T2. T3 

L10. LI 1.L 12 (TURNSCORE) 

(TURNS-CORE) 


19, No. 22 none 35, No. 24 25, No. 24 

T50-2 T68-2 T50-2, 2-t. link 

15, No. 22 none 20, No. 22 25. No. 24 

T50-2 T68-6 T50-2, 2-t. link 

12, No. 22 none 12, No. 22 28. No. 24 

T50-6 T68-6 T50-6, 3-t. link 

10, No. 22 21. No. 22 10. No. 22 19. No. 24 

T50-6 T50-6 T50-6 T50-6, 2-t. link 

Coil and transformer data. Toroid cores are Amidon Assoc, powdered-iron type. Y1, Y2, Y3 
and Y4 for 3.5 through 21 MHz, respectively, are 5.5. 5.2, 12.2 and 19.2 MHz (International 
Crystal Co. type GP, 30-pF load capacitance). 


7.0 to 7.2 

14.0 to 14.2 

21 .0 to 21 .2 


preceded by a bandpass preselector 
filter and followed with a diplexer and 
dual-gate MOSFET amplifier at 1.9 
MHz. A block diagram of the system is 
shown in Fig. 41. 

The original intention was to con- 
struct separate converters for each band, 
80 through 10 meters. However, after 
reviewing the design requirements, this 
was found to be redundant. Diode-ring 
mixers are inherently broadband and do 
not require tuned circuits. Furthermore, 
the post-mixer amplifier would be iden- 
tical for all of the bands. Only the 
front-end preselector networks and local 
oscillators need be changed between 
bands. The final configuration chosen 
was to use a master board which con- 
tained the diode-ring mixer and the post 
amp. A family of boards was then 
constructed, each containing a suitable 
local oscillator and the preselector net- 
work for the band of interest. 

Mixer and Post-Amplifier Board 

The circuit for the mixer and dual- 
gate MOSFET amplifier is shown in Fig. 
42. There are a few departures from the 
typical in this design. First, a diplexer is 
used between the mixer and the “post 
amp." A 2200-ohm resistor at the gate 
provides a termination, causing the 
mixer to see 50 ohms in the 1 .9-MHz 
frequency range. 

In order to simplify the band switch- 
ing. + 12 volts dc is supplied through the 
local oscillator port of the mixer. This is 


realized with an rf choke and suitable 
capacitors. 

The output of the amplifier was 
designed for broadband performance. 
To obtain a large bandwidth, the output 
transformer (Tl) is wound on a high- 
permeability ferrite toroid. A powdered- 
iron core should not be used for this 
transformer. It was found that a ferrite 
core with a permeability of 125 was not 
suitable in this position. Much better 
bandwidth and impedance matching was 
obtained with the core specified, which 
has a permeability of 2000. The 2200- 
ohm resistor in the drain circuit ensures 
that the output impedance presented by 
the amplifier is close to 50 ohms. This is 
important in order to assure that the 
input filters of the 160-meter receiver 
are properly terminated. 

A ferrite bead is used on gate 2 of 
the amplifier. This may not be necessary 
in some cases. However, it was included 
to lessen the possibility of uhf oscilla- 
tions occurring within the amplifier. A 
Fairchild FT-0601 or RCA 40673 dual- 
gate MOSFET can be used at Ql . 

Front-End Sections 

Shown in Fig. 43 is the circuit used as 
the front end for each of the lower-input 
bands (3 .5 -3 .7, 7.0-72 and 14.0-14.2 
MHz). Component values are given in 
Tables 1 and 2. 

The local oscillator for each of the 
converters uses a bipolar transistor and 
is designed to provide an output from 


+10 to +13 dBm. This level of L0 
injection was found to be near optimum 
for the diode-ring mixer. 

The preselector filters are fairly elab- 
orate. However, the results are well 
worth the extra expense and effort. 
Predistorted filter-synthesis methods 
were used when designing the bandpass 
filters. They were designed for a three- 
pole Butterworth response. 

One problem with multisection 
filters using capacitors as coupling ele- 
ments between the resonators is that the 
stop-band attenuation may degrade in 
the vhf spectrum. This is due to slight 
amounts of lead inductance in the tun- 
ing capacitors, and the fact that the 
capacitive-intersection coupling method 
degenerates toward a high-pass filter 
response away from the passband. In 
order to suppress these responses, 
should they occur, a 5-polc low-pass 
filter is included at the antenna ter- 
minal. 

Two methods were used for evalua- 
tion of the filter designs. First, after 
initial calculation of the component 
values, a computer program was used to 
determine the frequency response of the 
filters over a wide range. In this analysis, 
resistors were placed in the circuit to 
stimulate the distortion effects caused 
by the losses in the cores. 

After the filters were built and 
aligned in the home shop, they were 
checked with laboratory instrumenta- 
tion. In that case a Tektronix 7L13 
spectrum analyzer and TR-502 tracking 
generator were used. The measured re- 
sults around the passband corresponded 
with the computer simulation. The stop- 
band attenuation was measured, with 
one exception, to be over 100 dB for all 
three filters evaluated. The exception 
was for the 80-meter filter. At about 70 
MHz the attenuation degraded to about 
95 dB, but returned to the better values 
at frequencies up through 200 MHz. 

A Butterworth response was chosen 
because that filter shape is aligned easily 
with simple test equipment. Alignment 
is performed by driving the filter with a 
50-ohm signal generator and terminating 
the output in a sensitive 50-ohm detec- 
tor. The generator is set at the center 


Table 2 


BAND 

C4. C6 

C5, C20 

C7 

C8 

C9, C12 

CIO 

C11 

C13 

C14 

C16 

C17, C31 

Cl 8. C32 

C21 

(MHz) 

C19 (pF) ( P F) 

(pF) 

(pF) 

Cl 5 (pF) 

(pF) 

(pF) 

(pF) 

(pF) 

(pF) 

(pF) 

(pF) 

(pF) 

3.5 to 3.7 

790 

1580 

130 

— 

90 to 400 

12 



10 

_ 

91 

100 

400 

_ 

7.0 to 7.2 

450 

890 

43 

— 

90 to 400 

4.7 

— 

4.7 

— 

62 

100 

400 

— 

14.0 to 14.2 

220 

450 

33 

90 

20 to 90 

3.3 

90 

3.3 

90 

22 

47 

20 to 90 

_ 

21.0 to 21.2 

150 300. 

345 

— 

51 

20 to 90 

1.2 

51 

1.2 

51 

12 

47 

20 to 90 

20 to 90 


Fixed-value and trimmer capacitors. Fixed-value capacitors are silver-mica or similar high-O, stable types. Trimmers are mica compression 
type. See text for obtaining precise non-standard fixed-capacitance values. 


140 Chapter 6 



Fig. 44 — Diagram of the 15-meter front-end circuit. Numbered fixed-value capacitors are silver micas. Resistors are 1/2-W composition. See 
Tables 1 and 2 for other parts values. 


frequency of the filter and the variable 
capacitors are adjusted for a maximum 
response. Experimentally, it was not 
found necessary to readjust the filters 
when the swept instrumentation was 
available. 

The converter for the 15-meter band 
was built using the circuit in Fig. 44. On 
this band it was felt that a better noise 
figure might be useful. This was pro- 
vided by inserting an rf amplifier be- 
tween the low-pass filter and the band- 
pass circuit. The low-pass circuit was 
modified . The input section is a symme- 
trical pi network with a Q of 1 . This is 
followed by a pi network with a Q of 10 
and an impedance transformation from 
50 to 2000 ohms. A 3300-ohm resistor 
is used in the drain circuit to ensure 
proper termination of the bandpass 
filter. In the unit built, the drain was 
attached directly to the “hot” end of 
the resonator (L10). However, it would 
be desirable to reduce the gain some- 
what. This would be realized easily by 
tapping the drain down on the tuned 
circuit as shown. The terminating resis- 
tor should remain across L10. 

Those building the converter for 80 
meters may wish to also cover the 
75-meter phone band. While the filter 
shown could probably be realigned for a 
range about 100 kHz higher, the shape 
of the filter would no doubt deteriorate 
if it were moved farther. A better 
approach would be to change the value 
of the inductors. Proper results should 
be obtained by reducing the coils from 


35 to 32 turns, keeping all capacitor 
values the same. A 5.8-MHz crystal 
would be required for tuning the range 
from 4.0 to 3.8 MHz. 

Additional Design Notes 

The reader should note that the 
tuning will be “backward” for the 
80-meter band. This was done because a 
strong 1.7 -MHz local-oscillator signal 
would have appeared at the input to the 
post-mixer amplifier. This could have 
resulted in 1MD products. Furthermore, 
for the 75-meter band the crystal would 
have been at 2.0 MHz if low-side injec- 
tion were used. This would have placed 
a strong signal within the tuning range 
of the main receiver. If it is desirable 
that all hf bands tune in the same 
direction, the builder should pick high- 
side crystals for all of the bands. 

The approach used for the 15-meter 
converter in order to obtain low-noise 
performance could also be applied to 
the 10- and 6-meter bands. Filter de- 
signs for these bands can be extracted 
from the appendix. The image rejection 
might be a little poor with such a low i-f 
frequency in the 6-meter case. 

Another revision would be the con- 
struction of a high-performance 80- 
meter receiver with converters for the 
higher bands. The converters described 
would be suitable for this situation. The 
crystal frequencies would diange ac- 
cordingly. The diplexer between the 
diode mixer and the “post amp" should 
be redesigned. This could be done easily 


by halving the inductance and capaci- 
tance values used in the diplexer circuit. 
The broadband output circuit in the 
drain of Q1 should work equally well at 
3.5 MHz. The 15- and 20-meter band- 
pass filters were designed with enough 
bandwidth to cover the total band. This 
was done in order to keep the insertion 
losses at a reasonable level. A slightly 
wider filter would be required for the 
total 40-meter band. 

The converters are built on large 
circuit boards. This was done in order to 



Interior view of the converter unit. The 
boards are mounted edgewise. The mixer 
module is seen at dead center. A multi- 
section wafer switch, with shield partitions 
between wafers, should be used in place of 
the one seen in this photograph (see text). 




Advanced Receiver Concepts 141 



ensure a reasonable level of stopband 
rejection in the filters and to ease 
construction. Those interested in a more 
compact format should consider the 
inclusion of shields between the sections 
of the input bandpass filter and between 
the filter circuitry and the correspond- 
ing oscillators. It is useful to build 
miniature equipment when there is a 
need for small size. However, for high- 
performance home-station equipment, 
where considerable experimentation 
may be required, a larger format is often 
desirable . 

Because the pc boards shown in the 
photograph are quite large, the builder 
will probably elect to lay the circuits 
out fora more compact format. For this 
reason there are no pc-board templates 
and layouts available. 

Care should be taken when the 
front-end sections are band-switched. 


Shielding between switch wafers should 
have over 100 dB of isolation. Diode 
switching is not recommended unless 
the builder has equipment to evaluate 
the effects on IMD. The single-wafer 
switch shown in the photographs is not 
recommended. 

The only converter evaluated for 
IMD was the 14-MHz unit. Two-tone 
IMD measurements were performed and 
it was found that the output intercept 
of the converter was +22 dBm. This is 
more than sufficient for the application, 
since it greatly exceeds the input inter- 
cept of the 160-meter receiver, +7.S 
dBm. 

The gain and MDS were measured 
for all four converters. The signal genera- 
tor used was an HP-8640B. On the 
three lower bands, the noise figure was 
12 dB plus the loss of the input filters. 
Similarly, the gain of the converter was 


12.5 dB, minus the loss of the input 
filters. It was found that the gain and 
noise figure could both be improved by 
removing the 2200-ohm resistor at the 
gate of Q1 . There was a slight reduction 
in the output intercept, but not enough 
to cause problems. However, the low- 
pass part of the diplexer became much 
sharper in frequency response. This 
would make a front-panel trimmer con- 
trol necessary. 

The 15-meter converter performed 
differently. The net gain of this unit was 

32.5 dB and the noise figure was about 
3 dB. This is too much sensitivity to be 
usable at this frequency. It is recom- 
mended that the builder move the drain 
tap on the bandpass filter as outlined. 

The two-tone dynamic range of the 
complete receiver was measured at 88 
dB. Blocking occurred for an input over 
120 dB above the MDS. 


142 Chapter 6 



1 

Chapter 7 


Test Equipment and Accessories 


^Measurements are the key to ob- 
taining good results in amateur experi- 
mentation. This form of test procedure 
will help assure proper equipment per- 
formance while enabling the builder to 
establish a log of normal operating 
voltages and parameters. A laboratory 
logbook which contains such data will 
be useful when it becomes necessary to 
troubleshoot the homemade equipment. 
The information will be valuable when 
designing new circuits which employ 
some of the stages and devices used in 
earlier assemblies. 

Some amateurs have concluded that 
sophisticated and costly test equipment 
is needed to obtain high quality results. 
Certainly, this can be true if experi- 
mentation is taking place well within 
the state of the art. But, a lot of good 
work can be done with only a VOM. A 
great deal more can be achieved if the 
amateur is willing to construct some 
simple test equipment for his personal 
laboratory: A less than optimum mea- 


surement is still better than no measure- 
ment at all! 

From the foregoing commentary 
emerges a primary rule which the 
writers have adopted: Keep the test 
equipment simple! Another principle 
they have embraced is that of not 
planning so far ahead that every applica- 
tion conceivable shall be handled by the 
assortment of homemade test equip- 
ment. The more esoteric pieces of labo- 
ratory gear can be built on an as-needed 
basis. 

Some Basic Recommendations 

The number of power supplies 
needed in the workshop always seems to 
exceed the quantity available. For this 
reason it is best to utilize power supplies 
which are outboard from the test equip- 
ment. The exception might be in the - 
case of weak-signal sources which re- 
quire superb isolation to minimize un- 
wanted leakage. 

Dry -battery packs of various voltage 


levels are useful to the experimenter. 
They arc beneficial when it is necessary 
to effect a high degree of power supply 
isolation. Also, a variable-voltage regu- 
lated dc supply is extremely useful in 
the amateur laboratory. The circuit 
which illustrates the use of an LM317K 
1C ( Fig. 48) is suggested. 

Those who desire a high-current, 
ripple-free dc power supply may wish to 
consider inclusion of a 12-volt auto- 
motive battery in the shop. It can be 
"topped off’ by means of a trickle 
charger when it is not being used. The 
life span of such a battery can be 
increased by periodic high-current 
loading and recharging, say, two or 
three times a week. 

DC Voltage Measurements 

Ordinary VOMs (volt-ohm-milliam- 
meter) are suitable for much of the 
routine work done in the amateur lab. 
Some of the small imported instruments 
can be purchased for less money than 
one would spend to build a comparable 
tester from scratch. The primary limita- 
tion of most VOMs is, however, that of 
loading the circuit under test. A typical 
VOM will exhibit a characteristic of 
1000 to perhaps 5000 ohms per volt 
when applied to a circuit test point. 
Loading of this variety will sometimes 
cause incorrect readings (lower than 
normal). A more practical voltmeter is 
one which has a high input resistance, 
such as a VTVM (vacuum-tube volt- 
meter) or a solid-state equivalent. The 
latter can often be built at a cost lower 
than that of a factory-assembled unit or 
commercial kit. The complexity of a 
homemade instrument will depend upon 
the accuracy desired. Some practical 
examples follow. 

Low-Cost FET Voltmeter 

Fig. 1 shows a simple voltmeter 
which uses one active device - a JFET. 
It is designed to accommodate two 

Test Equipment and Accessories 143 



i 




Fig. 3 - Details of the rf probe for use with VTVMs or the circuit of Fig. 1 (see text). CR1 is a 
1N34A or equivalent. A 1 N91 4A silicon diode is suitable also. 


dc -voltage ranges, 0 to 2 and 0 to 20 
volts. For most amateur solid-state ex- 
perimentation it will not be necessary to 
measure dc levels greater than 20. The 
accuracy of this instrument is ample foi 
all but the most exacting applications 
(±10 percent). 

As the dc voltage at the gate of Q1 is 
increased, the FET current rises, causing 
an elevation in the voltage drop across 
source resistance R9. The level change is 
indicated at Ml, a 100-qA meter. Some 
current will flow in Ql even when no dc 
voltage is applied to the gate. Therefore, 
control R7 is adjusted to provide a zero 
reading on Ml. R8 is tweaked to pro- 
vide a full-scale meter reading when two 
volts of dc are applied through R4. It 
may be necessary to readjust R7 and R8 
a couple of times to effect final cali- 
bration. 

When the voltmeter is first turned on 
by means of SI, there may be a short 
stabilization period caused by internal 
changes in the FET (junction heating). 
For this reason it is best to calibrate the 
voltmeter after it has been turned on for 
approximately one minute. When it is 
used for voltage measurements later on, 
allow a one-minute warm-up period to 
assure proper zeroing of the meter. Fig. 
2 shows a circuit-board layout for the 
meter. Isolated pads have been formed 
by means of a Moto Tool and cutting 
bit. The builder may choose to mount 
R7 and R8 on the front panel of the 
tester case. This will permit recali- 
bration of the circuit as the battery 
depletes. For greatest accuracy, R1 
through R4. inclusive, should be 1- 


percent units. However, 5-percent re- 
sistors will suffice for most amateur 
work. 

Readout on Ml will be linear. That 
is, full-scale deflection will represent 2 
or 20 volts, depending on the range in 
use. Midscale readings will equal one 
and ten volts, respectively, and so on. 
The builder may find it helpful to draw 
a new meter scale, having two ranges 
represented 0 to 2, and 0 to 20 volts. 

Building an RF Probe 

Fig. 3 shows how an rf probe can be 
built for use with the voltmeter of Fig. 
I. It will be useful when determining 
relative rms values of rf voltage from 50 
kHz to at least 148 MHz. It can be used 
with numerous commercial VTVMs to 
provide accurate rms voltage measure- 
ments, provided the voltmeter with 
which it is used has a 10-megohm input 
characteristic. However, when employed 


with the circuit of Fig. 1 the readings 
will not be perfectly coincident with the 
calibration of the meter at Ml. The 
internal 4.7 megohm at Fig. 3 is chosen 
to change the peak rf voltage response 
of the probe to an rms value compatible 
with voltmeters which have the 
10-megohm characteristic. 

Despite the lack of accuracy re- 
sulting from utilizing the probe with the 
circuit at Fig. I, signal tracing and 
relative rf voltage readings can be taken 
during circuit development or trouble- 
shooting. When used with a 10-megohm 
instrument, best accuracy will result 
when the waveform under test is a pure 
sine wave. Distorted waveforms will 
change the voltage readings signifi- 
cantly. 

The probe is made from a short 
length of copper tubing (3/8 or 1/2 inch 
in diameter). Wooden end plugs are 
installed to fit snugly inside the tubing. 
The probe tip can be made from a small 
nail or a piece of brazing rod which has 
been sharpened to a point on one end. 

Op-Amp Voltmeter 

Shown in Fig. 4 is a simple voltmeter 
that uses a pair of op-amp ICs and a 0-1 
mA meter. Type 741 op amps may be 
used. A better choice would be the 
LM-308N. This unit has the advantage 
of requiring low power from the battery 
and has low bias currents, leading to 
better accuracy. If the LM-308N is used, 
a 1000-pF capacitor should be con- 
nected between pins I and 8 of the chip 
in order to provide stable frequency 
compensation. 

In this circuit U1 serves as a fed-back 
current amplifier. Two input resistors 
are selected with a slide switch to 
provide full-scale readings of 2 and 20 
volts. The gain of the circuit is 0.5, 
leading to a 1 -volt change at pin 6 of U 1 
for a full-scale reading. U2 is used to 
provide a synthetic ground. This allows 
the circuit to be powered from a single, 
9-volt battery of the kind used in 
transistorized bc-band receivers. 

A pair of diodes is provided at the 
input to protect the semiconductors 
from excessive input voltages. The two 



144 Chapter 7 





1N914 4 

J1 

20V 



Fig. 4 Circuit of the op-amp voltmeter. 


controls in the circuit serve to calibrate 
the meter movement and to zero the 
output when there is no input signal. 

This meter functions like a VOM 
with a sensitivity of 500 k£2 per volt. 
The input resistance changes for the 
different ranges. Because of this, the 
circuit cannot be used with the usual rf 
probe. Most rf probes are built to work 
with a VTVM or FET voltmeter that has 
a constant input resistance of 10 meg- 
ohms. As in Fig. 3, they usually contain 
a 4.7 megohm resistor. Such a probe 
could be used with good accuracy on 
the 20-volt range of the meter in Fig. 4, 
but errors would occur on the 2-volt 
scale. 

Shown in Fig. 5 is another FET 
voltmeter. This circuit is the semicon- 
ductor equivalent of some popular 
VTVMs. A dual FET is used in this 
circuit, resulting in exceptionally low 
drift characteristics with temperature 
changes. Also, the FET chosen is a unit 
with a low pinchoff voltage. This has 
the asset that the meter may be 


powered from a low-voltage supply. The 
required 6 volts are provided by means 
of four D-type dry cells. Since the 
current consumption is only a few mA, 
Pen-light cells would serve as well. The 
circuit is a full differential amplifier. 
Each side consists of the FET and a pnp 
transistor arranged as a noninverting 
amplifier with feedback to produce a 
voltage gain of 2. The output of this 
amplifier is applied to an emitter fol- 
lower to drive the meter. 

The dual JFET used in the schematic 
may be a difficult item to obtain. 
However, if the voltage is increased in 
the circuit, almost any dual FET will 
work. If a dual FET cannot be located, 
the modified amplifier shown in Fig. 6 
is recommended, where individual FETs 
of the same type are used. Two units of 
similar characteristics should be chosen. 
They should be matched for I dss and 
pinchoff voltage. 

The unit utilizes a meter with a 0-1 
mA movement, but with three scales 
labeled 0-70, 0-140 and 0-350. The 


resistive divider was designed specifi- 
cally to be compatible with these scales, 
with a circuit sensitivity of 0.35 volt full 
scale. In the circuit shown in Fig. 5, the 
basic sensitivity is assumed to be 0.5- 
volt full scale, and the resistive divider 
has been designed to yield full-scale 
sensitivities of 0.5, 1, 2, 5, 10, 20, 50, 
1 00, 200 and 500 volts. The sensitivity 
is controlled with the range switch, SI. 
A double-pole, double-throw slide 
switch, S2, is used for polarity reversal, 
while S3 serves to switch power to the 
meter. Although not shown in the 
schematic, a second set of contacts on 
S3 is arranged to short out the meter 
movement when the unit is off. This is a 
good practice with high quality meter 
movements to prevent damage during 
transit. 

In the modified circuit of Fig. 6, the 
pair of FETs are used as source followers 
to drive a pair of 741 op amps. The 
741s then drive the meter. This circuit 
could use either a 747 or a 5558 dual op 
amp in place of the two 741s. While the 



Fig. 5 A semiconductor equivalent of some of the popular VTVMs used bv amateurs. The circuit exhibits low-drift characteristics respective to 
temperature changes. Q1 • Dual N-channel JFET, Vp * 1.5V. 

Test Equipment and Accessories 145 










drift of this circuit is certain to be 
greater than when a dual JFET is used, 
it should still be better than those 
circuits which contain a single FET. 

RF Power Measurement 

One of the most frequent measure- 
ments performed by the amateur experi- 
menter is that of rf power. The most 
common application is during the 
testing of transmitters. The receiver 
builder needs to know the power avail- 
able from his TO and BFO. Also, if he is 
to evaluate the dynamic range of his 
receiver, he must have signal generators 
with known output powers. These are 
obtained with low-power oscillators fol- 
lowed by a step attenuator. The output 
power must be measured before applica- 
tion of the attenuator. 

For hf transmitter work, rf power is 
most easily measured with a high-level 
diode detector and a dummy load or 
termination. A circuit suitable for 
powers of 10 or 15 watts for short time 
periods is shown in Fig. 7. Six 300-ohm, 
2-W resistors have been paralleled to 
serve as the termination, Rl. Detection 
is performed with a 1N914 diode, and 
the dc voltage is monitored with a 
voltmeter. Any VOM is suitable at the 
higher power levels. 



Solid-state voltmeter which uses FETs. 


146 Chapter 7 


The diode serves as a peak detector. 
That is, the largest positive voltage 
appearing across the 50-ohm termina- 
tion is the value that the capacitor 
attains, and is measured by the volt- 
meter. For a sine-wave input, which is 
the usual waveform of interest, the 
power is given as P = V 7 d( . -f 2 R where 
R is the termination, in this case equal 
to 50 ohms. 

As higher powers are to be mea- 
sured. simple techniques like those 
shown in Fig. 7 may not be suitable. 
The reason is that the peak reverse 
voltage appearing across the diode may 
exceed the diode breakdown specifica- 
tion. One simple way of circumventing 
this problem is shown in Fig. 8 where a 
voltage divider is placed across the 
termination. The net termination should 
still equal 50 ohms. The measured volt- 
age must be multiplied by the appro- 
priate division factor in order to calculate 
the power with the previous equation. 
With voltage-divider techniques, the 
power-measuring capability is easily ex- 
tended to the 1-kW level. 

Significant errors appear when the 
methods of Fig. 7 are extended to low 
powers. The major source of error is the 
V-I characteristic of the diode. Recall 
that a silicon diode like the 1N914 



Fig. 7 — A high-level diode detector (or rf 
power measurements into a dummy load. 
See text for information concerning Rl. 


requires about 0.6 to 0.7 volt across it 
before significant current flows. Hence, 
with rf powers corresponding to a peak 
voltage of 0.6 volt, no detected output 
will appear. (Actually, there may be 
some, but the accuracy of the measure- 
ment will be poor.) 

The first step toward better sensi- 
tivity is to substitute a more sensitive 
diode type. Either a germanium or a 
hot-carrier silicon diode would be a 
much better choice, since they turn on 
at much lower voltages. Values for 
diode turn-on voltage down to 0.1 
to 0.2 volts are common. For best 
accuracy the voltmeter should draw 
minimal current from the detector. 
Hence, a VTVM or FET voltmeter is 
preferred over a simple voltmeter. 

Shown in Fig. 9 is a power meter 
that is built on the back of a 500-/aA 
meter. This unit uses a hot-carrier diode 
detector and will yield an indication for 
input powers as low as +1 or +2 dBm. 
The resistor was chosen for a full-scale 
reading of + 1 7 dBm (50 milliwatts). 

A meter of this type cannot be used 
to determine power with a simple for- 
mula. The reason is that the value of the 
diode offset voltage is too close to the 
peak rf voltages being measured, leading 
to excessive errors. However, meters of 



Fig. 8 — A voltage divider can be employed to 
increase the power measuring capability of the 
circuit shown in Fig. 7. 










Rf power meter seen assembled on the back 
of a meter. 


Fig. 9 Circuit of an rf power meter which 
can be built on the back of a 500-pA meter. 



ion 



Fig. 1 1 - Circuit for proper biasing to obtain square-law detection. 


this kind are easily calibrated by noting 
that the circuit is still a peak-reading 
detector. This allows a dc calibration to 
be done. 

Imagine that a power of 1 0 inW was 
to be measured. This power would 
correspond to 1-volt peak across a 
50-ohm resistor. To calibrate the meter 
for 10 mW, place 1-volt dc across the 
termination and note the meter re- 
sponse. Similarly, 2-volts dc would 
correspond to 40 mW. Using this 
method, a calibration curve can be 
generated for the power meter. In the 
unit shown, such a calibration was 
found to correspond within 1 dB of 
that from industrial instrumentation. 

While a sensitivity near 1 mW is 
adequate for most situations, it is often 
useful to be able to measure powers 
which are much lower. One approach to 
this would be to precede the diode 
detector with a broadband amplifier. A 



Fig. 10 — Small-signal waveform applied to a 
diode detector and the resultant output. 


better approach, however, is to increase 
the basic detector sensitivity before 
adding amplifiers. The simplest way to 
do this is by biasing the diode detector 
with dc. 

Shown in Fig. 10 is a small-signal 
waveform applied to a diode detector 
and the resulting output. Note that an 
input voltage as small as that shown 
(about 0.1 -volt peak) would produce no 
current in a diode with zero bias. 
However, when the voltage is applied to 
the biased diode, we see a definite 
current flow. The current that flows is 
not what we would expect if the diode 
were replaced with a resistor. Instead, 
we see that the positive-going half of the 
input voltage yields a much larger 
current flow than the negative part. The 
result is that if the diode current is 
monitored, a dc component is present. 
This form of detection is usually re- 
ferred to as "square law” detection. The 
mathematics are outlined in the ap- 
pendix under a discussion of distortion 
phenomena. 

In order to achieve square-law 
action, a diode must be biased carefully. 
Specifically, it should be biased at a 
constant current level from a low im- 
pedance dc source. While this could be 
achieved with a battery and a variable 
resistor, a much better method is to use 
an operational amplifier. 

Shown in Fig. 1 1 is a circuit to 
accomplish this biasing. A pair of iden- 
tical diodes are used. However, only one 
(C R1 ) has rf applied. The other serves as 
a reference for properly biasing the 
detector. With this circuit, input powers 
as low as -26 dBm (3 microwatts) can 
be detected. 

The calibration is straightforward. 
An oscillator is built to deliver about 


+ 10 dBm output. This- power is easily 
measured with the peak detector de- 
scribed earlier. The oscillator output is 
applied to a step attenuator with up to a 
40-dB range. The available output 
powers are now suitable for the square- 
law detector, and are well defined 
within the errors of the collection of 
instruments. 

The diode square-law detector is 
quite flat from about 1 MHz up through 
the vhf spectrum. Either hot-carrier 
diodes or small-signal silicon switching 
diodes can be used. If better op amps 
were used with lower drift specification. 



Fig. 1 2 - Diagram of a broadband amplifier 
which can be used to extend power. meter 
sensitivity to lower power levels. T1 con- 
tains 7 bifil.ar turns of enameled wire on an 
Amidon FT-23-43 toroid core. Circuit gam 
is 19 dB and the bandwidth is 175 MHz. 


Test Equipment and Accessories 147 
















EXCEPT AS INDICATED, DECIMAL 
VALUES OF CAPACITANCE ARE 
IN MICROFARADS I^Fl; OTHERS 
ARE IN PICOFARADS I pF OR j,>F); 
RESISTANCES ARE IN OHMS', 
h -IOOO. M- 1 000000 I f 


r^ [ 


^r jf 




01 

2N3179 1 


-r< OUTPUT 
I (SO OHMS) 







i2o> ;=:.i 



91^ -T-.l 


Fig. 1 3 — A four-stage broadband rf amplifier. Gain = 40 dB and the upper 3-dB point of the amplifier is 65 MHz. 


[he system could be operated with 
higher dc gain, yielding even better 
sensitivity. Some manufacturers make 
diodes which will detect signals down to 
-50 dBm. 

The best way to extend sensitivity to 
lower power levels is with a broadband 
amplifier. Shown in Fig. 12 is a single- 
stage amplifier using a 2N5179. Heavy 
feedback is used to stabilize gain and to 
provide 50-ohm input and output im- 



J)/ 


Exterior of the broadband. 50-ohm amplifier. 


. “‘ o Ju *■ i d "‘id 

*** • J - - ■>. j 

~.-v — v • 


Interior layout of the broadband amplifier. 

148 Chapter 7 


pedances. The 3-dB points in this circuit 
were about 2 MHz and 175 MHz. 

The 50-ohm transducer gain was 19 
dB, the noise figure 6.5 dB (at 10 MHz), 
and the output intercept +24 dBm. Gain 
compression starts near +10 dBm. 

Shown in Fig. 13 is a four-stage 
amplifier. The upper 3-dB point in this 
amplifier was about 65 MHz and the 
gain was 40 dB. Noise figure was not 
measured. 

These amplifiers are useful acces- 
sories for applications other than 
power measurements. For example, 
they may be used as preamplifiers for a 
frequency counter, or even a receiver. 

Shown in Fig. 14 is a block diagram 
of a useful general-purpose instrument. 
An attenuator, amplifier and sensitive 
detector are combined for a wide sensi- 
tivity range. If the input is driven from 
an outboard tuned circuit, a wave meter 
of spectacular sensitivity would result. 

In-Line RF Power Measurement 

Rf power measurements can be 
made accurately at specified impedance 


levels by using an rf bridge circuit of the 
type illustrated in Fig. 15. The basic 
circuit was described by Bruene in QST 
for April, 1959. The concept was 
treated in a practical manner by DeMaw 
in QST for Dec., 1969. 

The principle of operation is that the 
inner conductor of a coaxial trans- 
mission line passes through the center of 
toroidal transformer T1 to function as 
the transformer primary. A multiturn 
secondary winding is placed on the core. 
Rf current through the primary induces 
a voltage in the secondary, causing 
current to flow through R1 and R2. The 
voltage drops across these resistors are 
equal in amplitude, but are 180 degrees 
out of phase with respect to common, 
or ground. Practically speaking, they are 
in and out of phase, respectively, with 
the line current. Capacitive voltage divid- 
ers. CI/C3 and C2/C4, are connected 
across the line to secure equal-amplitude 
voltages in phase with the line voltage. 
The division ratio is adjusted so that 
these voltages match the voltage drops 
across R1 and R2 in amplitude. These 


/-hi (o- 


SOUARE-LAW 

DETECTOR 


ATTENUATOR 
(0-60 OBI 


Fig. 1 4 — This block diagram illustrates a test instrument which contains an attenuator, 
amplifier and sensitive detector. 



conditions exist at only a specified load 
impedance - usually 50 or 75 ohms to 
match the characteristics of the trans- 
mission line. Initial adjustment of the 
bridge is done while using a resistive 
load standard of the value desired. 

Under the foregoing conditions, the 
voltages rectified by CR1 and CR2 
represent, in one case, vector sum of the 
voltages caused by the line current and 
voltage. In the other case, the vector 
difference is represented. With respect 
to the resistance for which the circuit 
has been adjusted, the sum is propor- 


tional to the forward component of a 
traveling wave of the variety that occurs 
on a transmission line, and the differ- 
ence is proportional to the reflected 
component. 

Fig. 15A shows the main portion of 
the power bridge as being contained in a 
shielded enclosure, as indicated by the 
dashed lines. External to the shield are 
the components needed to meter the 
forward and reflected components. In 
the example at A, a single potentio- 
meter is used to set the full-scale power 
indication of Ml. In this case R3 can be 


calibrated for various full-scale power 
levels by observing the rms output 
voltage from the bridge with an rf 
probe, or the pk-pk value by means of a 
scope. The voltage is measured across a 
resistive termination which matches the 
characteristic impedance of the bridge 
unit. A 10-turn Helipot and mating dial 
mechanism will allow greater reset ac- 
curacy than will a simple control-and- 
knob arrangement. 

Fig. 15B shows an alternative tech- 
nique for presetting the instrument for a 
specific full-scale power level. Trimpots 
can be mounted inside the instrument 
case and adjusted for a particular power 
sensitivity; e.g., 10. 50, 100, 500 or 
1000 watts. If more than one power 
range is desired, an assortment of con- 
trols can be used, then switch-selected 
for the power ranges required. 

It is important to maintain good 
isolation between the through-line 
ports, and between the line and the 
remainder of the bridge circuit. It is 
good practice to use an isolating divider 
such as that seen in the photograph of 
Fig. 16. Some manufacturers who fol- 
low this general design, utilize a Faraday 
screen between the primary and second- 
ary windings of T1 . This helps prevent 
unwanted capacitive coupling, thereby 
aiding the nulling of the bridge circuit. 

The bridge is balanced by connecting 
a 50-ohm signal source to the input 
port, and terminating the output port in 
50 ohms, resistive. With the instrument 
set to read reflected energy, Cl is 
adjusted for a zero reading at Ml. The 
load and source cables are reversed next, 
and the procedure repeated while 
adjusting C2 for a zero meter reading. 
Following the null adjustments the 
builder can calibrate the instrument for 
a specific full-scale power level, as dis- 
cussed earlier in this treatment. Bridges 
of this general type are suitable for use 



Fig. 16 - Photograph which shows a shield 
divider between the rf and dc portions of the 
bridge (double-sided pc-board strip across 
center of box). 



Fig. 15 - Schematic diagram of an rf power bridge. T1 has 60 turns of no. 30 enameled wire 
and uses an Amidon T68-2 toroid core. Cl and C2 should be piston or air trimmers to assure 
alow minimum capacitance. CR1 and CR2 can be 1N34A or 1N914A diodes (matched pair 
recommended). See text for a discussion of the circuits at A and B. 


Test Equipment and Accessories 149 





Fig. 17 - Schematic diagram of a QRP rf power meter. It is suitable for levels from 1 to 100 
watts, 1 .8 to 30 MHz. T1 contains 60 turns of No. 30 enameled wire and uses a T68-2 toroid 
core. The primary of T1 consists of two turns of No. 20 insulated wire. Cl and C2 follow the 
rule set forth for the circuit of Fig. 15. 


lip to 30 MHz. The lower frequency 
limit, with the component values given, 
is approximately 1.8 MHz. If a pc-board 
fomiat is used, the constructor may 
elect to employ pc-board strip-line tech- 
niques to assure a relatively constant 
50-ohm line characteristic between the 
input and output ports. The value of 
such an approach will be seen at 21 
MHz and higher, where the composite 
bridge can cause a slight line-impedance 
discontinuity (a line “bump”) if the 
through-line is not close to 50 ohms. 

If a 50-/rA meter is used at Ml. 
maximum forward-power sensitivity for 
this circuit will be on the order of 10 
watts. This type of bridge is not “fre- 
quency conscious,” as is the Monimatch 
circuit popularized in QST. That is, it 
will respond uniformly to a given power 
level from 1.8 to 30 MHz. Nulling 
adjustments should be done at the 
highest frequency of use (30 MHz in 
this example). 

A QRP Power Meter 

Fig. 17 illustrates a suitable bridge 
for use in measuring power levels from 1 
to 100 watts. The circuit is a variation 
of that shown in Fig. 15. To increase 
the sensitivity, a two-turn link is used 
for the primary. This represents a slight 
tradeoff in through-line impedance at 
the higher end of the hf spectrum, but 
the line discontinuity is not great 
enough in magnitude to spoil the utility 
of the instrument. 

Figs. 18 and 19 show the construc- 
tion technique used. R1 has been cali- 


brated for a full-scale reading a! Ml of 5 
watts. The calibration chart atop the 
bridge case shows power levels from 
0.25 to 5 watts, versus the meter-scale 
markings. Phono jacks and SO-239 type 
connectors are connected in parallel at 
the input and output ports, purely for 
utility. 

Attenuators 

An attenuator is one of the most 
useful accessories that the amateur can 
have in his shop. It will allow a given 
power source to be reduced by a known 
factor. If the amount is variable, as 
would be the case with a step attenua- 
tor, the unit can be used with a sensitive 
power-measuring meter in order to 
determine gain and to evaluate linearity. 
Attenuators may be used to extend the 
range of existing sensitive power meters 
to arbitrarily high levels. 

High quality attenuators are avail- 
able commercially and are fairly expen- 
sive. Alternatively, step attenuators may 
be constructed from resistors and slide 
switches. While the accuracy is certainly 
not as good as one would realize with 
better units, it is usually sufficient for 
amateur work. Again, we offer that a 
measurement of less than optimum 
precision is better than no measure- 
ment ! 

Attenuators have assets other than 
reducing the power in a controlled way. 
Since they are made from resistors, they 
will change a source or load that may be 
highly reactive into one that is known 
and resistive. Similarly, a source or load 



Fig. 18 - Exterior view of the QRP power 
meter. A 4 X 4 X 2-inch aluminum utility box 
serves as a case. Phono jacks are placed in 
parallel with uhf connectors at the input and 
output ports of the unit to permit use of a 
greater variety of cable connectors. A cali- 
bration mark can be seen at the left center 
of the meter. The mark represents the con- 
trol setting for 5 watts full scale. 

that is of unknown impedance may be 
made to appear as a clean, resistive 
termination with an attenuator. The 
extent to which these effects occur will 
depend upon the amount of attenuation 
employed the more attenuation, the 
more closely the load approaches the 
characteristic impedance of the attenua- 
tor. 

There are a number of circuits that 
may be used to form resistive attenua- 
tors. Three of these are shown in Fig. 20 
along with the appropriate design equa- 
tions for choosing resistor values. These 
equations are derived easily from first 
principles if the experimenter is so 
inclined. There are two vital conditions 
that must be satisfied. First, the power 
delivered to the load must be a known 
ratio of that supplied to the input of the 
attenuator. Second, the input resistance 
seen at one end of the attenuator should 
equal the desired characteristic resis- 
tance, R 0 , when the output is termina- 
ted in the same value. Using these 


Fig. 19 — Interior view of the power meter of 
Fig. 17. 



150 Chapter 7 






Fig. 20 — Three circuits for forming resistive 
attenuators. 



Fig. 21 — Circuit for a step attenuator which 
s useful into the vhf spectrum. 



Outside view ot the step attenuator. 


Table 1 


5-Percent Resistor Values for Simplo 
Attenuators 



7T 


T 


L 


A.dB r 

R 

r 

R 

r 

R 

1 

910 

6.2 

2.7 

390 

5.6 

390 

2 

430 

12 

5.6 

220 

10 

200 

3 

300 

18 

9.1 

150 

15 

120 

6 

150 

39 

16 

62 

24 

51 

10 

91 

68 

27 

36 

33 

24 

20 

62 

240 

39 

10 

43 

5. 


condilions, ihe equations may be set up 
so that, when solved, they yield the 
design equations shown. 

When using the equations in Fig. 20, 
A is the attenuation ratio in dB. The 
voltage attenuation ratio, "e," is related 
to A with the equation given in the 
figure. 

Care should be used in the construc- 
tion of attenuators with slide switches. 
If 1 -percent tolerance resistors are avail- 
able, they should be used. However, the 
results are often quite suitable with 5- 
percent resistors. Kvery effort should be 
made to keep the lead lengths as short 
as possible. This will help to extend the 
upper frequency of usefulness. Shields 
are beneficial if the unit is to be used at 
vhf. This is especially significant for 
single sections of 20 dB or more. 

Three types of attenuator are 
shown: a pi, a T and an L circuit. The 
pi and the T are symmetrical, and are, 
thus, the more useful types. The L 
circuit has the problem that the output 
resistance of the section may be much 
different than the input resistance of 
Ihe circuit. In some cases, this presents 
no obstacle. For switchable attenuators, 
the pi circuit offers the best compatibil- 
ity with the slide switches. A circuit for 
a step attenuator is shown in Fig. 21. 
The photograph shows a unit that offers 
good accuracy up through the vhf spec- 
trum. 

Shown in Table 1 is a list of values 
of common 5-percent resistors that may 
be used for various amounts of attenua- 
tion. Half- or quarter-watt resistors are 
suitable for small-signal work. For high- 
er power units, the specific circuit must 
be evaluated carefully to ascertain the 
power specifications of the resistors. As 
an example, consider Ihe 10-dB pi atten- 
uator shown in Fig. 22, and assume that 
it is to be designed for a resistance of 50 
ohms. Assume that the maximum input 
power, when properly terminated, will 
be 10 watts, which corresponds to a 
voltage of 22.4 across 50 ohms. 

Solving the equations given earlier, 
the resistor values are 96.3 ohms at the 
ends and 71.2 ohms for the connecting 
arm. If we solve for the voltages, which 
are shown in circles in Fig. 22, we may 
calculate the powers dissipated in the 
three resistors. The input resistor dissi- 



pates 5.19 watts, the 71.2-ohm resistor 
consumes 3.29 watts, while the output 
resistor consumes only 0.52 watt. A 
good choice for the input resistor would 
be a parallel combination of two 300- 
ohm ones and a 270-ohm unit, all with a 
2-watt dissipation rating. The connect- 
ing arm could be another parallel pair of 
2-watt units with resistances of 150 and 
130 ohms. The output could be a 
I -watt, 91 -ohm resistor. If such an 
attenuator was built for rf power mea- 
surement, the input should be clearly 
marked. 

The attenuators discussed here have 
been dissipative devices, with some of 
the input power applied to them being 
absorbed within the circuit. However, 
other methods are useful for measure- 
ment applications that are not dissipa- 
tive. Shown in Fig. 23 is one example, a 
20-dB coupler. This is a high- 
pernieability ferrite toroid core set up as 
a current transformer. The primary of 
the transformer is a single wire passing 
through the core while Ihe secondary is 
a 10-turn winding. If the secondary is 
terminated in a 50-ohm load, such as a 
low-level power meter, this termination 
will reflect back through the trans- 
former according to the square of the 
turns ratio. Hence, the core will appear 
as a 0.5-ohm resistor in series with the line. 
If the main line is also terminated in 50 
ohms, the net resistance presented to 
the source is 50.5 ohms (essentially 
unchanged). Noting that the ratio of the 
two resistances is 100, or 20 dB. the 
power delivered to the power meter will 
be attenuated from that delivered to the 
main load by 20 dB. Techniques of this 
kind can be applied to the evaluation of 
higher power sources (such as trans- 



Fig. 23 - An example of a 20-dB coupler. T1 
uses a single mire through the toroid core as 
the primary. The secondary is a 10-turn mind- 
ing of enameled mire. An FT-23-43 core is 
used. 


Test Equipment and Accessories 151 







Fig. 24 - A Wheatstone bridge for measuring 
dc resistance. 



Fig. 25 — An alternative to the circuit of 
Fig. 24. 



mitters) when being evaluated with low- 
power instrumentation. Note that this 
unit is not a directional coupler - it 
makes no difference which way the 
current is flowing. 

Bridges for RF Measurements 

A useful instrument is an rf bridge. 
While the classic application of such a 
device is for antenna and transmitter 
evaluation and tuning, there are a num- 
ber of other applications. Most of the 
measurements done with bridges occur 
at relatively high-power levels. However, 
often one wants to determine the im- 
pedance of low-power active circuits. If 
the usual high-level bridges were used in 
measuring such circuits, the results 
would be inaccurate. In the extreme, 
the circuit being studied could be dam- 
aged. 

Consider the Wheatstone bridge that 
is used for dc resistance measurements. 
This is shown in Fig. 24. Assume that 
voltage E is applied to the bridge, and 
that resistors R1 and R2 are equal in 
value. This being true, the voltage at 
point A will be El 2. The other two 
resistors in the bridge are R s , a “stan- 
dard,” and R x , the unknown resistance. 
The voltage at point B will be deter- 
mined by the ratio of the two resistors. 
If R s and R x are equal, the voltage at 



Fig. 27 — A bridge circuit which has a sensi- 
tivity control. 


point B will also be E/2. The bridge is 
now balanced and there is no voltage 
difference between point A and B. 
Thus, there will be no indication in the 
detector. 

What will happen in the more typical 
case where R s and R x are not equal? 
Since the voltage at point B is no longer 
£'/2, a potential difference exists be- 
tween points A and B and a current will 
flow in the detector. We could calibrate 
the meter to tell us the level of un- 
balance, and thus infer the value of the 
unknown resistance, R x . However, a 
better approach would be to make the 
standard, R s , a calibrated variable resis- 
tor. It could then be varied until a null 
is indicated with no response in the 
detector. Then knowing R s . and observ- 
ing that the bridge is balanced, we know 
the value of R x . 

Shown in Fig. 25 is another 
approach. Here, we have replaced R1 
and R2 with a potentiometer. R s now 
has a fixed value. The control is varied 
until a null is again achieved. A bridge 
of this kind is calibrated by placing 
various known values in the R x posi- 
tion. The dial on the control is then 
marked accordingly. 

The foregoing examples occurred at 
dc, so the detector would be a meter 
with a capability for deflection in either 
direction (zero center). However, the 
same principles will apply if a different 
kind of detector is used and the input 
driving voltage, £', is an rf sine wave. 
Such a bridge is shown in Fig. 26. The 
resistors are all 50 ohms. However, for 
the bridge to operate properly, this is 
not necessary. The only requirement is 
that R1 and R2 be equal, and R s is the 
same as the load the bridge is designed 
to measure. The typical values for R s 
are 50 or 75 ohms. 

The detector in the rf bridge is a 
diode in series with a capacitor. Assume 
that the unknown impedance is a 50- 
ohm resistor. In this case the bridge will 
be balanced because the rf voltages at 
points A and B are equal. There is no 
potential difference across the detector. 


Consider now the case where a 100-ohm 
resistor is placed across the R x port. 
The voltage at B will be higher than that 
at A. This voltage difference will appear 
across the detector diodd and will 
charge the capacitor to some dc voltage. 
This will cause a current to flow 
through the 10-kf2 resistor and the 
meter, giving an indication. A similar 
result would occur if a 25-ohm resistor 
were placed on the unknown terminal. 

Consider now the case where the 
unknown impedance had a magnitude 
of 50 ohms, but was reactive. For 
example, the unknown load could be a 
35-ohm resistor in series with an induc- 
tor that had 35 ohms of reactance at the 
input frequency. The bridge would not 
be balanced. While the magnitudes of 
the impedances are proper to balance 
the bridge, the fact that the unknown 
termination is reactive means that the 
voltages at points A and B are not in 
phase with each other. An analysis will 
show that this leads to a detector 
output. In order for the bridge to be 
balanced, the unknown load must be 50 
ohms and be purely resistive. 



Exterior view of the bridge. The small unit is 
the return-loss bridge of Fig. 36. 


152 Chapter 7 


l 








Inside view of the bridge. Note short leads in 
Ihe signal path. 


The bridge just described is useful in 
spite of its simplicity. Shown in Fig. 27 
is the circuit of a similar unit that has a 
potentiometer added as a sensitivity 
control. The unit is shown in a photo- 
graph. By keeping the leads short, and 
by using a germanium diode, the bridge 
is reasonably accurate through the 2- 
meter band. It can be driven with as 
little as 100 mW of rf power. The small 
size makes it convenient for rooftop 
adjustment of antennas. 

Shown in Fig. 28 is a similar unit 
using a control for the ratio arm of the 



Fig. 28 — Here a control is used as the ratio 
irm of a bridge. 


bridge. This unit is useful for experi- 
mental work since a wide variety of 
resistances can be measured, ranging 
from, say, 10 to 1000 ohms. In a bridge 
of this kind the exact value of the 
“standard” resistor is not critical, for 
this will merely determine the R x value 
for which the control will be in the 
center. The bridge is calibrated by sub- 
stitution of known resistances at the R x 
port. The major limitation of this instru- 
ment is its upper frequency limit. This 
arises from the capacitance of the amt 
of the control to ground. The reactance 
will be constant (more or less), but the 
resistance above the arm of the control 
will vary, leading to a variable phase for 
the reference voltage of the bridge. 

The problem of errors from stray 
capacitances can be circumvented by 
replacing the variable resistance arm 
with a variable capac'tance voltage divid- 
er (Fig. 29). It may be shown that such 
a divider produces a voltage that is in 
phase with the driving signal. Sevick, 
W2FMI, has described several bridges of 
this kind (see the bibliography). The 
advantage is that stray capacitances are 
absorbed in the variable element and do 
not lead to frequency-dependent errors. 

All of the bridges described have the 
capability of measuring only resistances. 
If a reactive termination is present, a 
complete null cannot be obtained. How- 
ever, reactive impedances may be mea- 
sured by using an outboard adaptor as 
shown in Fig. 30. This unit is a series- 
tuned circuit. The inductor is chosen so 
the bridge will see a null when a resistive 
termination is placed on the output and 
the variable capacitor is at midrange. In 
practice, the capacitor and the resis- 
tance-measuring arm in the basic bridge 
are adjusted repeatedly until a complete 
null is obtained. The position of the 
variable capacitor in the reactance- 
canceling arm will tell the user if the 
termination is inductive or capacitive. 
The system may be calibrated if desired. 

Bridges for Antenna Tuners 

Consider now a bridge that might be 
used to tune a Transmatch. Such a unit 
is shown in Fig. 31. This bridge differs 
slightly from the others we have con- 
sidered: A resistor has been added at the 
input, and the values of the resistors in 
the divider arm have been reduced from 
50 to 15 ohms. These changes are 
significant. Consider the impedance ex- 
tremes that can appear at the output 
termination. One is a short circuit, while 
the other is an open circuit. For these 
two extremes, the resistance seen at the 
input of the bridge will vary only from 
46 to 57 ohms. Both values are close to 
50 ohms. As a result, the transmitter 
will always see something close to a 
proper termination. This can be a pro- 
found advantage if the transmitter being 
used to drive the bridge is prone to 



Fig. 29 - A variable-capacitance voltage 
divider is used in this circuit to replace a 
resistive divider. 



Fig. 30 - An outboard adapter for use in 
measuring reactive impedances. 



Fig. 31 — A bridge circuit suitable for use 
when adjusting a Transmatch. 



Fig. 32 — A high-power adaptation of the 
circuit shown in Fig. 31. 


Test Equipment and Accessories 153 











Fig. 33 - A capacitive voltage divider in 
parallel with a transmission line. 


self-destruction when a mismatch 
occurs. Severe mismatches can occur 
during the tuning of a Transmatch. An 
additional advantage of the bridge 
shown is that, when matched, the out- 
put applied to the antenna is down 12.8 
dB from the full transmitter output that 
is applied to the input. Use of bridges of 
this type would help eliminate carriers 
during tune-up periods. 

This absorptive-bridge technique is 
by no means limited to low power 
applications even though the unit of 
Fig. 31 can be driven with less than a 
watt. Shown in Fig. 32 is a high power 
adaptation of this method. One of the 
writers has used this technique when 
tuning the station Transmatch for 
several years. It’s comforting to know 
that only 50 mW of rf is reaching the 
antenna during tune-up periods even 
though 25 or 30 watts is available from 
the transmitter. 

In many cases, a bridge of the kind 
described above is not sufficient. In- 
stead, a unit that operates at full power 
is desired. Such units are useful for 
monitoring antenna VSWR on a contin- 
uous basis, or for measuring the input 
VSWR of a high-power amplifier. The 
latter could vary as a function of drive 
power. 

In the section on attenuators earlier 
in this chapter, a ferrite transformer was 
used as a 20-dB coupler. In this applica- 
tion, the voltage appearing across the 
coil secondary was proportional to the 
current (lowing in the line. Consider 
now the effect of a capacitive voltage 
divider across the transmission line (Fig. 



Fig. 34 - Illustration of a 2M8 coupler in 
combination with a capacitive voltage 
divider. 


33). The voltage at point A will be in 
phase with the voltage on the line. 
However, the magnitude of the voltage 
will be one-tenth the value on the line. 

Consider the result of combining the 
two effects. This is shown in Fig. 34. 
The voltage appearing across the termin- 
ating resistor,/?,, is proportional to the 
current flowing in the transmission line. 
The voltage appearing from the capaci- 
tive divider is proportional to the volt- 
age on the line. The ratio of these two 
quantities, E -s- /, is indicative of an 
impedance. Assume that the capacitors 
are adjusted such that the voltage from 
A is the same magnitude as the voltage 
across R,. Then, when the connection is 
made at point X in the circuit, the two 
voltages will add in phase. The resultant 
will be detected by the diode, producing 
a dc output. 

Consider now the effect of reversing 
the in-line bridge. That is, the port that 
was terminated with the 50-ohm load is 
now driven by the transmitter, and the 
original input is terminated in 50 ohms. 
The voltage at point A will be virtually 
the same. However, the current is now 
flowing in the opposite direction from 
the earlier case. Because of this, the 
voltage appearing across R, will be out 
of phase by 180 degrees from the 
original case. The two rf voltages will 
now cancel each other. No detected 
output will occur. Units of this type are 
appropriately called directional bridges. 

In the typical unit, a double second- 
ary is used on the transformer in order 
to allow both forward and reverse 
powers to be monitored simultaneously. 
Some examples are seen in Figs. 15 and 
17. 

The Return-Loss Bridge 

Let us return now to a simple 
resistive bridge. Shown in Fig. 35 is a 
bridge that departs slightly from those 
described earlier. First, it is driven from 
a 50-ohm source. This was not neces- 
sarily the case when a transmitter was 
used. The output impedance of a trans- 
mitter could look like something very 
much different than 50 ohms, even 
though it may have been designed to be 
terminated in a 50-ohm load. The 
second difference is that a 50-ohm 
resistor is connected between points A 
and B. Clearly, if R x is 50 ohms, the 
bridge is balanced and there is no 
voltage difference between points A and 
B. There will be no power dissipated in 
the detector resistance, R d . 

Assume that the unknown port is 
now either open or short circuited. It 
may be shown that in either of these 
cases an identical voltage difference will 
appear across R d . If the. bridge is not 
driven from a 50-ohm source, the volt- 
age across R d will not be the same when 
one goes from a short to an open 
circuit. 



Fig. 35 — Another version of a simple resis- 
tive bridge. 


The circuit has a drawback. Most 
50-ohm detectors (like those described 
earlier in this chapter) are single-ended. 
This deficiency may be solved with the 
circuit of Fig. 36, where a “sortabalun" 
has been inserted from the floating 
detector port to a single-ended port. 
This allows the voltage difference be- 
tween points A and B to appear across a 
single-ended output. Also, the imped- 
ance presented to the single-ended 
detector port is now impressed between 
points A and B. The transformer has 
approximately 10 bifilar turns of No. 30 
enameled wire on an FT -23 43 ferrite 
toroid. Ferrite should be used instead of 
powdered iron. 

When using the bridge, the unknown 
port is either short or open circuited, 
and the power in the detector is noted. 
Then, the unknown termination is 
attached to the unknown Z port and the 
detector power is again noted. The 
ratio, expressed in dB, is known as the 
return loss. The higher the return loss, 
the closer the unknown termination is 
to 50 ohms. It may be shown that the 
return loss ( R-L ) is related to the 
magnitude of the reflection coefficient 
r , by R-L - 20 Iog| </. The reflection 
coefficient is related to the voltage 
standing wave ratio by r = (VSWR - 1) 
t (VSWR + 1). Table 2 compares 
return loss, reflection coefficient and 
VSWR for a wide range of values. If 
phase angle is to be included, a more 
complete representation would be r = 


so 



Fig. 36 — A return-loss bridge tor impedance 
measurements. See text. 


154 Chapter 7 












Table 2 

/ 



YSWR = —L 

T • 


1 

- r 

RETURN 

r , 


LOSS. 

REFLECTION 


dB 

COEF. 

VSWR 

1 

0.891 

17.4 

2 

0.794 

8.72 

3 

0.707 

5.85 

4 

0.631 

4.42 

5 

0.562 

3.57 

6 

0.501 

3.01 

7 

0.447 

2.61 

8 

0.398 

2.32 

9 

0.355 

2.10 

10 

0.316 

1.92 

12 

0.251 

1.67 

14 

0.199 

1.50 

16 

0.158 

1.38 

18 

0.126 

1.29 

20 

0.100 

1.22 

26 

0.056 

1.12 

30 

0.032 

1.07 

35 

0.018 

1.04 

40 

0.01 

1.02 

45 

5.6 X 10 3 

1.011 

50 

3.16 X 10/ 

1.006 

60 

1.0 X 10 3 

1.002 


(Z Z a ) -s- (Z + Zq). All of these 
parameters are of significance when 
using a Smith chart for impedance 
representations. 

One major advantage of a return-loss 
bridge is that the measurement of imped- 
ance can be done at low-power levels. 
For example, a low-level signal genera- 
tor could be used as the rf source, and 
one of the sensitive rf detector systems 
described earlier could be used as the 
detector. In fact, a receiver could be 
used in conjunction with a step attenua- 
tor as the detector. The simple detec- 
tors described will provide only infor- 
mation about the magnitude of the 
reflection coefficient. To measure the 
angle, a vector voltmeter would be 
needed. 

Another application of the return- 
loss bridge would be as a simple 6-dB 
hybrid combiner. A typical application 
would be to combine the outputs of 
two signal generators for the purpose of 
measuring intermodulation distortion 
and gain compression in, for example, a 
receiver. One generator is applied to the 
source port while the other is connected 
to the detector port. Shown in Fig. 37 is 
such an application. Assuming that each 
generator is set to deliver 10 mV to a 
50-ohm load, the resulting voltages are 
shown. Note that generator A delivers 5 
mV to the output load, hence the 6-dB 
loss. However, note that 5 mV appears 
at both of the detector points in the 
bridge as a result of drive from genera- 
tor A. There is no voltage difference, 
and none of the signal from generator A 
appears at generator B. The converse is 
also true. This is needed in IMD mea- 
surements. If one generator is allowed to 
"talk to the other,” the result may be 
that one generator will phase modulate 


the other. This modulation leads to 
sidebands at the same frequencies where 
IMD products appear and can cause 
errors in the IMD measurements. 

Solid-State Power Supplies 

Nearly all of the equipment in this 
book requires an external dc power 
source. Although some battery-powered 
gear is described for field use, the 
subject of batteries shall not be treated 
here. Rather, we will focus attention on 
power supplies and voltage regulators 
which operate from the ac power line. 
Some rules of thumb are offered for 
those who wish to design and build their 
own power supplies and regulators. A 
more concise treatment of the general 
subject can be found in The Radio 
Amateur's Handbook, and in the refer- 
ences given in the bibliography section. 

A Basic Power Supply 

Fig. 38 shows a typical unregulated 
dc power supply. A quad of silicon 
rectifier diodes is used in a full-wave 
hookup. Since full-wave bridge rectifica- 
tion is the most efficient of the com- 
mon types, we shall deal with that 
circuit in this chapter. 

An advantage of a bridge rectifier is 
that it delivers full-wave output without 
the need for a transformer with a 
secondary center tap. Another feature 
of the full-wave rectifier is that the 
ripple frequency at the output is twice 
the line frequency, thereby making fil- 
tering less difficult. Thus, the capaci- 
tance of the filter capacitor for a speci- 
fied percentage of output ripple will be 
considerably lower than with a half- 
wave rectifier. 

A Design Example 

Let’s assume we need a simple power 
supply that is able to provide a voltage 
output of 13. Maximum current taken 
by the external load will be 500 mA 
(0.5 A). Maximum ripple will be 3 



Fig. 37 — A 6-dB hybrid combiner can be used 
to connect two signal generators to a test cir- 
cuit for measuring, as one example, receiver 
dynamic range. 


percent, and the load regulation will be 
5 percent. 

The rms secondary voltage for Tl of 
Fig. 38 must be the desired V a plus the 
voltage drops across CR2 and CR4 (= 
1 .4 V) divided by 1 .41 . Thus, Tl V, ec = 
13 + 1.4/1.41 = 10.2 volts. The nearest 
standard transformer would be a 10-volt 
one, which would be close enough in 
value. Alternatively, the builder could 
wind his own transformer, or remove 
secondary turns from a 12-volt trans- 
former to obtain the desired rms second- 
ary voltage. 

A 3-percent ripple referenced to 13 
volts is 0.39 V rms. Therefore, the 
pk-pk value is found from: V rip = 0.39 
X 2.82 = 1 .09 V. This figure is necessary 
to calculate the required capacitance for 
Cl. 

Also needed for determining the 
value of Cl is the time interval (/) 
between the full-wave rectifier pulses, 


CR2 



V„ (no load) s V„ c X 1.41 

Po = V a X 1 1. 

Ri. = V 0 *I l 



Cl (F m ,„)= Vsec x 1.41 

FI (A) = 21 IN (N = turns ratio) 

V*c*V 0 *\A\ 


Fig. 38 A circuit which illustrates the configuration of a basic unregulated dc power supply. . 

Test Equipment and Accessories 155 







Fig. 39 — Zener diodes are effective as simple 
voltage regulators. 

which for that circuit is 8.3 milliseconds 
(ms). Therefore, Cl is calculated from 

ClOtF) = l ^~ 

v rip 

[b.5A X 8.3 X ltr 3 

L 1.09 
= .0038 X 1 0 6 = 3800 /tF 

(Eq.l) 

where Ii = the current taken by the 
circuit which is powered by the supply 
V 0 . The nearest standard capacitor 
value is 4000 */F. It will be an accept- 
able one to use, but since the tolerance 
of electrolytic capacitors is rather loose, 
a 5000-pF unit will probably assure that 
the design requirements are met. 

Diodes CR1-CR4. inclusive, should 
have a PRV rating of at least two times 
V aec peak, which means with our ex- 
ample we have 14.4 volts. Therefore the 
PRV should be 28.8 or greater. Four 
50-V diodes will work nicely. Similarly, 
the forward current of the diodes (l/ w d) 
should be at least twice the load cur- 
rent, So for a 500-mA //. the diodes 
should be rated at 1 A or greater. 

The load resistance, R,, is deter- 
mined by V 0 /I l , which in this example 
is 13/0.5 = 26 ohms. This factor must 



be known in order to find the necessaiy 
series resistance for the target 5-percent 
regulation. R>imax) = Load reg. X 
R L I 10 = .05(26/10] = 0.13 ohm. 
Therefore, the transformer secondary dc 
resistance should be no greater than 
0.13 ohm. The secondary current rating 
should be equal to or greater than the 
//, of 0.5 A. A transformer of that type 
will usually have a secondary resistance 
of less than our maximum acceptable 
amount for a 5-percent regulation trait. 

Information on calculating the value 
of the fuse, FI, is given in Fig. 38. Cl 
should have a minimum working voltage 
of 18.33 in accordance with the formula 
in Fig. 38. The next standard value is 
suggested a 25-volt capacitor. 

Regulated Voltages 

When the need arises to regulate 
small amounts of current, say. up to 
100 mA, Zener diodes offer a low-cost 
approach. Even though higher amounts 
of current are handled sometimes by 
Zener diodes, the practice is not a 
common one in amateur work. Our 
treatment will be confined to the lower 
current amounts. 

Most Zener diodes are known alsoias 
avalanche diodes. They are similar tn 
construction to junction rectifiers, but 
the primary characteristic for their in- 
tended purpose is the reverse- 
breakdown profile. In simple terms, 
positive voltage is applied to the cath- 
ode of the diode rather than to the 
anode. As this reverse voltage is made 
higher the leakage current in the diode 
stays fairly constant until a critical 
plateau is reached. This point is known 
as the breakdown voltage. There is a 
marked contrast between the end result 
of the breakdown point of a Zener 
diode and a conventional rectifier diode. 
With the latter it is essential to operate 
the diode well below the breakdown or 
PRV (peak reverse voltage) to avoid 
damaging it. When the breakdown point 
of a diode is reached, copious amounts 
of current flow through the junction, 
and in the case of Zener diodes this area 
is known as the Zener current. 

At breakdown, the normal high back 
resistance of the diode drops to a very 
low amount and, therefore, the current 
increases rapidly. The amount of cur- 
rent is, however, limited by the series 
resistance (R, of Fig. 39) between the 
diode and the voltage source. The rated 
breakdown value of a Zener diode is 
that level for which the semiconductor 
was designed. Typically, the plateaus 
range from 3.9 to as high as 200 volts. 
The amount of safe sustained Zener 
current is determined by the wattage 
rating of the component. These values 
run from 150 mW to 50 watts at 
present. 

Because of the characteristics we 
have just described it can be seen that a 


Zener diode will serve nicely as a voltage 
regulator, sine-wave clipper, or as a 
series-gate element. Voltage regulation is 
made possible by virtue of the high 
current which flows at conduction. The 
regulator current must always be con- 
siderably higher than that which is 
drawn by the l L (circuit to which the 
regulated voltage is applied). Under that 
rule the significant current which flows 
through the series dropping resistor is 
that of the diode: Small changes in 
input voltage or circuit load current are 
disguised by the diode current and R s 
by means of the E = / X R rule. 

Designing with Zener Diodes 

There are three sets of conditions 
common to regulator circuits: variable 
load current and constant supply volt- 
age, constant load current and variable 
supply voltage, and variable load current 
and variable supply voltage. A slightly 
different equation applies in each case. 
Figs. 39 and 40. 

A rule of thumb can be used with 
respect to the ratio of minimum Zener- 
diode current ( Izmin ) and the load 
current (//,). For best regulation the ratio 
should be 10: 1 . That is, the load current 
should be roughly 10 percent of the 
Zener diode current. 

Fig. 39 shows a shunt type of 
Zener-diode regulator. It provides 9.1 
volts regulated to a VFO which has a 
constant load current of 10 mA (.01 A). 
The 10:1 current ratio does not result 
from the values given, but the figure is 
close enough for most amateur work. 
Had a lower value of V z been chosen. 


Fig. 40 — Zener-diode application for circuits 
which have changes in load current. 



156 Chapter 7 






V sec ( rms ) ss 1 .4 V a 
Cl (pF) - See Eq. I 
Cl (F)=*2k" 
C2{V mln )>V c 


C2 GiF) - 0.5 Cl C//F) F* = V (uc ,(rms) X 1 .41 
Rp ^ V„ X 80 
VR\ = F„ +0.7 
v o-—V z - 0.7 


^ = K„ X I L 
Rl = Vo+Il 
F\ = /,. X 2 


Fig. 41 - Illustration of a power supply with regulation, A pass transistor, Q1, is used to extend 
the range of the Zener -diode regulator. 


the ratio would have been closer to the 
suggested one. 

In the equation of Fig 39A a series 
resistance of 173 ohms is u . i. The 
nearest standard value is 180 ohms. 
That will be entirely suitable for R s . 

The equation at Fig. 39B determines 
that the maximum Zener-diodc power 
dissipation is 0.167 W. A good rule of 
thumb for choosing a wattage rating for 
the diode is a times-5 factor. This will 
allow ample safety margin for diode 
internal heating. Since we determined 
that VR1 will dissipate 0.167 W, a 
5-times value will be 0.8 W. The nearest 
standard power value is 1 W, so a diode 
of that type will suffice. 

Fig. 39C gives an equation for com- 
puting the wattage dissipated in R, at 
Yintinax/’ which is0.138 W. To stay on 
the safe side of things we will again use 
the 5-times rule. This gives us a wattage 
rating for R, of 0.69. In practice, a 
1/2-watt resistor will suffice that 
being the nearest standard value. 

When high-wattage Zener diodes 
must be used (10- to 50-W types, in 
general), they will be of the stud-mount 
variety. Heat sinking is done in the same 
manner as with power transistors and 
power-type rectifier diodes. The general 
rules for this have been given earlier in 
the book. A more complete discussion 
of Zener-diode applications was given in 
QST for April, 1976. 

Extending Zener-Diode Range 

The foregoing section outlines some 
of the limitations when using Zener 
diodes as regulators. Greater current 
amounts can be accommodated if the 
Zener diode is used as a reference at low 
current, permitting the bulk of the I L to 
flow through a series pass transistor (Ql 
of Fig. 41). An additional benefit in 
using a pass transistor is that of reduced 
V 0 ripple. This technique is sometimes 
referred to as "electronic filtering.” 

Q1 of Fig. 41 can be thought of as a 
simple emitter-follower dc amplifier. It 
increases the load resistance seen by the 
Zener diode by a factor of beta (0). In 


this situation VR1 is required to supply 
only the base current of Ql. The net 
result is that the load regulation and 
ripple characteristics are improved by a 
factor of beta. Addition of C2 reduces 
the ripple even more, although many 
simple regulated power supplies of the 
type seen in Fig. 41 do not have C2 as a 
part of the circuit. 

The primary limitation of this type 
of circuit is that Ql can be destroyed 
almost immediately if a severe overload 
occurs at R B . The fuse, FI , cannot blow 
fast enough to protect Ql. Further- 
more, if a low-current fuse was used at 
Vo it would be subject to the same 
limitations. In order to assure longevity 
of Ql it is necessary to include a 
current-limiting circuit of the kind 
shown in Fig. 42. Modern three-terminal 
regulators have replaced the circuit of 
Fig. 42, and that subject will be dis- 
cussed later in the chapter. 

It should be mentioned that the 
greater the value of V, ec at Tl, the 
higher the power dissipation in Ql . This 
not only reduces the overall efficiency 
of the power supply, but requires strin- 
gent heat sinking at Ql. The circuit of 
Fig. 41 could be made to operate with a 
V, ec as great as 25 volts, but a more 
suitable voltage level for a 13-volt out- 
put at V„ would be 18 volts rms. In this 
regard it is not difficult to remove the 
required number of secondary turns 
from a 24-volt transformer. 

A Design Example 

We desire a regulated, well-filtered 
dc voltage of 13. f maximum shall 
be 0.5 A. The circuit of Fig. 41 will be 
the one used in this example. The ratings 
lor Tl, CR1-CR4, and Cl can be deter- 
mined by using the formulas given for 
the circuit of Fig. 38. V svc shall be 18 
V rms. 

In order to calculate the value of)?, 
in Fig. 41 we must learn what l b (base 
current) for Ql will be. The base cur- 
rent is approximately equal to the 
emitter current of Ql in amperes 
divided by beta: J„ = 0.5/25 = .02A. or 


20 mA. The transistor beta can be 
found in the manufacturer’s data sheet, 
or measured with simple lest equipment 
(beta = l c /I b ). Since the beta spread for 
a particular type of transistor 2N3055 
for example, where it is specified as 20 
to 70 - is a fairly unknown quantity, 
more precise calculations for Fig. 41 
will result if the transistor beta is tested 
before the calculations are done. A 
suitable, conservative approach is to 
design for beta minimum of the transis- 
tor used. 

As we learned earlier, in order for 
VR1 to regulate properly it is necessary 
that a fair portion of the current flow- 
ing through R s should be drawn by 
VRI. Therefore, let us set a rough rule 
of 30 mA for 1r Knowing this figure, 
plus the //, of .02A just computed, the 
Zener-diode current (/,) will be .03 A 
.02 A = .01 A, or , J mA. From this we 
can learn that: /?,(ohms) = (V' 
V z )IIr, = (25.3 14)/ .03 = 376 ohms. 

The nearest standard ohmic value for /?, 
is 390, so it shall be used. The wattage 
ratings of /?, and VRI can be obtained 
from the formulas given earlier for 
Zener-diode regulators. 

A safe power rating must be pro- 
vided for Ql. In this context it should 
be known that the dissipation in Ql will 
be equal to the emitter current times 
the collector-to-emitter voltage. Thus, 
for our circuit of Fig. 41 P QI = l E X 
V CE , where V CE equals the desired V' 

(V 2 V BE ). Therefore, Pq 1 = 0.5 A 
X 12 V = 6 watts. V BE for a silicon 
transistor is approximately 0.7 V. A 
good rule of thumb in this example is to 
choose a transistor at Ql which has a 
p D(nu ix) of at least twice Pqi There- 
fore, Ql should be rated at 12 watts or 
more. Since the cost of power transis- 
tors is quite low, a 25-, 50-, or 100-watt 
unit will allow considerable safety fac- 
tor if heat-sinked properly, and would 
represent a good choice. 

Load regulation with the power 
supply of Fig. 41 will be approximately 
2 percent, and the output ripple will be 
low. Line regulation will be on the order 
of 7 percent, assuming the 117-V line 
has variations. 

The .01 -pF capacitors at the primary 
of Tl serve two functions. They act as 
transient suppressors and help prevent rf 
energy from entering the power-supply 
regulator. C3 serves in a similar manner. 
R p is used as a minimum-load resistance 
for periods when the power supply has 
no external load. 

Current Limiting 

Damage to Ql of Fig. 41 can occur 
when the lj exceeds the safe amount, 
or when I becomes excessive. Fig. 42 
illustrates a simple current-limiter cir- 
cuit which will protect Ql . All of the //, 
passes through R2. Therefore a voltage 
difference will exist across R2, the 


Test Equipment and Accessories 157 






Fig. 42 — Overload protection for a regulated dc supply can be effected by addition of a current- 
overload protective circuit contrasted to that of Fig. 43. 


precise amount being dependent upon 
the exact //_ value at a given time. When 
the load current exceeds a predeter- 
mined safe value, the voltage drop 
across R2 forward biases Q2 and causes 
it to conduct. Since CR5 is a silicon 
diode, and because Q2 is a silicon 
transistor, the combined voltage drops 
through them (roughly 0.7 V each) will 
be 1 .4 V. Therefore, the drop across R2 
must exceed 1.4 V before Q2 can turn 
on. This being the case, R2 is chosen for 
a value that provides a drop of 1 .4 V 
when li/max) occurs. In this instance 
1 .4 volts will be seen when I E reaches 
0.5 A. 

When Q2 turns on, some of the 
current through R, flows through Q2, 
thereby depriving Q1 of some of its base 
current. This action, depending upon 
the amount of Q1 base current at a 
precise moment, cuts off Q1 conduction 
to some degree, thus limiting the (low 
of current through it. 

Specifications 

Addition of the current limiter will 
cause a loss of roughly 1 .4 volts over 
that obtained from the circuit of Fig. 
41, owing to the inclusion of R2. 
Therefore, if a V„ of 13 is desired, the 
output from Q1 should be 14.4 V. 

Q2 can be a medium-beta, low- 



Fig. 43 - An Improved current-overload 
protective circuit contrasted to that of Fig. 
42. 


power device. It must be able to sustain 
the full V„. In this example a 25-V V CE 
will be ample, and a Pp of 1 W will be 
suitable for 02. A 2N2102 would be a 
good choice. 

R1 will be approximately 100 times 
the Ri value. Since R/, in this example 
is 26 ohms, V 0 llL(max)< Rl W >H be 
2,600 ohms. The value of Rl can be 
trimmed to provide Q1 cutoff when /, 
exceeds the safe amount. 

R2 is chosen from R2 = 1 .4 V/0.5 
A = 2.8 ohms. The closest standard 
resistor value is 3 ohms, which should 
be acceptable. R2 must handle l Umax) 
without overheating. Therefore, its dissi- 
pation will be 0.5* X 3 = 0.75 W. A 
2-watt resistor should allow sufficient 
safety margin. Magnet wire of small 
cross-sectional area can be used to wind 
R2 . This practice will enable the builder 
to obtain the precise ohmic value 
needed. 

Refinements in Discrete Regulators 

In the example of Fig. 42, suitable 
performance was obtained for the case 
where a constant load current was to be 
supplied. The ripple of the power 
supply was fairly low and the output 
voltage was reasonably stable. However, 
there are some inexpensive refinements 
that may be applied to simple regulators 
which will improve performance signifi- 
cantly. 

The first thing that can be done to 
improve regulation is to decrease the 
resistance value of Rl (Fig. 42). In the 
circuit shown the design was tailored 
such that a 1.4 volt drop would occur 
across R2 when the output current was 
0.5 A. However, if the load was re- 
moved from the output, the voltage 
would go from the desired output level 
of 13 volts up to 14.4 volts. Q2 is not 
turned on until the power supply goes 
into current limiting. 

The diode in the regulator circuit 


(CR5) provides a well-defined current 
where limiting occurs. However, if the 
desire is mainly to protect tire power 
supply from self-destruction, this diode 
may be eliminated, as may Rl. The 
result is shown in Fig. 43. This circuit 
has better load regulation. At full cur- 
rent (0.5 A) the output voltage is 13. 
When the load is removed, the voltage 
goes up to 13.7. Note that it was 
necessary to decrease the value of the 
Zener diode from 15.1 to 14.4 volts. 

While this is a 2:1 improvement in 
regulation over that originally obtained, 
it is still less than desired for many 
situations. Another problem is that the 
exact value of the Zener diode has a 
direct bearing on the output voltage 
obtained. The typical voltage tolerance 
of inexpensive Zener diodes is ±5 per- 
cent. A 5 percent variation in the 
14.4-volt diode required could allow the 
output to range from 13.7 down to 12.3 
volts. A more desirable situation would 
be a power supply that used a lower 
voltage Zener diode and an additional 
transistor. The exact output voltage 
could then be set with a variable resis- 
tor. Such a power supply regulator is 
shown in Fig. 44. We will assume that 
the Zener diode chosen has a rating of 
6.2 volts. 

When power is initially applied to 
this circuit, the series pass transistor is 
turned on with the 300-ohm bias resis- 
tor. This causes the voltage at the 
output to increase in value. The output 
voltage is attenuated by resistors Rl and 
R2, and causes a voltage to appear at 
die base of Q2. This turns transistor Q2 
on, and charges capacitor Cl. Cl will 
charge until it reaches the Zener-diode 
voltage of VR1. The Zener diode then 
clamps the voltage at the emitter of Q2 
at 6.2 volts. The base voltage on Q2 will 
be 0.7 volt greater, or 6.9 volts. 

What will the output voltage be? 
Assume that the two resistors are equal 
in value and that their ohmic value is 



Fig. 44 - This regulator circuit is more 
precise than that of Fig. 43, permitting the 
builder to obtain a specific output voltage. 


158 Chapter 7 







reasonably low. The low value ensures 
that the current flowing in the resistors 
is large in comparison with the base 
current in Q2. Since the resistors are of 
equal value, the voltage at the junction 
of the two resistors must equal 0.5 the 
output voltage. But, because of the 
Zener diode and the e-b drop of Q2, the 
base voltage at Q2 must be 6.9. Hence, 
the output must be equal to twice this 
value, or 13.8 volts. 

The foregoing analysis was carried 
out for no external load on the power 
supply. What happens if a resistive load 
is new placed on the supply? This 
would tend to drop the output voltage. 
However, when this begins to occur, the 
voltage on the base of Q2 will decrease. 
As this happens, the collector current in 
Q2 will decrease also. This will cause a 
reduced voltage drop across R,. This 
means that the voltage on the base of 
Q1 will increase, causing the output 
voltage to again increase until it reaches 
13.8. The voltage drop across the 1.4- 
ohm current-limit sensing resistor has no 
effect upon the output voltage. 

This voltage regulator utilizes an 
amplifier in a negative feedback loop. 
The fact that the output voltage was not 
affected by the drop across the 1 .4-ohm 
sensing resistor was the result of the 
feedback signal being obtained after the 
current limiting circuitry. The limiting 
circuit (Q3) was within the feedback 
loop. 

If the desired output was not 13.8 
volts, but 13 volts as before, it could be 
obtained by changing the ratio of R1 to 
R2: If R1 were 470 ohms, the output 
voltage would be 13.0 volts when R2 
was 4 1 5 .5 ohms. The best way to design 
this power supply would be to make R2 
a 500-ohm variable resistor. Then tire 
output could be adjusted from 6.9 to 
14.2 volts. 


♦VIUNREG.) O' Rl- 3.2 



I Rfl. 45 - Example of some refined tech- 
niques for use in a regulated power supply. 


The current limiting still functions. 
If the output current becomes high 
enough that 0.7 volt is developed across 
the sensing resistor, Q3 will turn on. 
This will then rob base current from Q1 , 
the pass transistor. The output voltage 
will then decrease accordingly, with no 
more than 0.5 A flowing in the external 
load. When the power supply is short 
circuited (crowbarred), the current will 
remain at 0.5 A. 

Shown in Fig. 45 is a regulator that 
demonstrates some refined techniques 
that might be used in a regulated power 
supply. In looking back at the regulator 
of Fig. 44, we see that Q2 functioned as 
an inverting amplifier. It can be shown 
that the regulation of the circuit is 
directly dependent upon the gain in this 
amplifier. In the supply of Fig. 45, we 
have replaced the single, discrete transis- 
tor amplifier with a high-gain opera- 
tional amplifier. A 741 will function 
well in circuits of this kind. However, a 
741 has a maximum output current of 
around 10 mA. This would not have 
been enough to drive the base of Q1 
directly if high output currents were 
desired. Hence, another transistor, Q2, 
is added to form a Darlington-connected 
pass transistor. The effective beta of 
such a configuration is approximately 
the square of die beta of a single 
transistor. It is reasonable to assume an 
effective beta for the combination of 
500 to 1000. Because of diis high beta 
value, die op amp needs to deliver only 
a few mA of current to the base of Q2 
for an emitter current in Q1 of 1 
ampere. 

The current limiting is different in 
diis circuit than it was in Fig. 44. Note 
that die emitter of Q 3 is tied to the 
output directly. However, the base of 
Q3 is biased from a voltage divider from 
die current sensing resistor. This divider 
has a ratio of 5/6. That is, the voltage 
on die base of Q3 is (5/6) V ei , where 
V el is the voltage on die emitter of Ql. 
Let’s assume that the regulator is to go 
into current limiting when die load 
current reaches 1 A. Widi the emitter of 
Q3 at the output voltage of 12, the base 
voltage must be equal to 12.7 at this 
instant. Due to the divider action, the 
voltage on die emitter of Ql, the pass 
transistor must be (6/5)12.7 = 15.2 
volt: We choose a sensing resistor of 3.2 
ohms. 

This circuit has tremendous implica- 
tions when we consider the behavior of 
the supply under a crowbar condition. 
With the output shorted, the emitter of 
Q3 is at 0 volts, and the base will be at 
0.7 volt. Following the earlier analysis, 
the emitter of the pass transistor will be 
at 1 2 times diis level, or 0.84 volt. The 
current in the supply is then 0.84/3.2 
ohms = 0.26 amperes. This is much less 
than the current that die supply will 
deliver prior to going into limiting. This 


technique is called fold-back current 
limiting. The advantage is that the 
supply components need not be capable 
of handling such high currents during 
short-circuit conditions. 

The price to be paid for this 
extreme protection is that die unregu- 
lated voltage must be higher. This is 
because there will be higher voltage 
drop across the sampling resistor, Rl 
prior to the point where limiting occurs. 

Another feature of die regulator of 
Fig. 45 is the nature of die reference 
diode biasing. The reference is a 6-volt 
Zener diode which is biased to a current 
of about 13 mA. The diode establishes 
the bias on the noninverting input of 
die error amplifier. The output voltage 
is established by adjusdng R2 . The asset 
of biasing the Zener diode as shown is 
diat virtually all of die current in the 
Zener comes from the regulated output. 
In earlier supplies, such as diat shown in 
Fig. 43, the Zener diode is biased from 
the unregulated supply which has high 
ripple. Measures of diis kind will help 
immensely in removing die last traces of 
hum from a power supply output. 

If a builder is construcdng power 
supplies using the techniques outlined in 
Fig. 45, care must be exercised to 
ensure that device specificadons are not 
exceeded. Specifically, the maximum 
supply voltage rating of a 741 op-amp is 
30 volts between pins 7 and 4. Since pin 
7 is connected to die unregulated 
supply, this value should not exceed 30 
volts. 

Somedmes it is desirable to build 
variable voltage supplies that will go all 
the way down to 0 volts. This can be 
done with a modification of the circuit of 
Fig. 45. A negative power supply is 
first built and is well regulated. A 
typical value might be -6 volts. This 
supply is used to provide operating 
voltage for pin 4 of die 741. Pin 3 of 
die 74 1 is grounded directly. The end of 
Rl, which is presently grounded, is 
returned to the negative supply. 

Three-Terminal Regulators 

Power supply design has been simpli- 
fied in recent years by the appearance 
of the three-terminal regulator 1C. These 
units contain all of the essential com- 
ponents for voltage regulation and cur- 
rent limiting. These include a high-gain 
error amplifier, sensing resistors and 
transistors for current limiting, a 
temperature-compensated voltage refer- 
ence, and suitable pass transistors. These 
ICs are available in a number of fixed- 
voltage ratings from 5 to 24. They may 
be obtained for load currents up to 3 
amperes, and come in various package 
styles. 

These ICs have a number of advan- 
tages. The main one is die simplicity of 
application. The three terminals are for 
a ground reference, an input for the 

Test Equipment and Accessories 159 









Fig. 46 - The illustration at A is that of a 12-V, 0.5-A supply which employs an LM341-12 reg- 
ulator 1C. Shown at B is a resistive divider which permits elevating the 1C output voltage above 
the value it is designed to handle (see text). 



(A) 


(B) 


Fig. 47 - Method of extending the current range of a regulator 1C. Here we see Q1 , a pass 
transistor, "wrapped around" U1 to increase the current capability of the power supply. 


JA CR2 



BOTTOM VIEW 


Fig. 48 - Circuit of a continuously variable regulated supply which utilizes the LM317K 
regulator 1C. CFtl through CR4 are 100-PRV, 3-A diodes. Line regulation is .01 percent/V and 
load regulation is 0.1 percent. 


unregulated voltage and an output. The 
ground reference is usually connected to 
the mounting surface. Because of this.it 
is not necessary that the IC be electri- 
cally insulated from ground. This eases 
the heat-sinking problem. Another typi- 
cal feature is that of “thermal shut- 
down." If the chip should become 
excessively warm due to insufficient 
heat sinking, the temperature rise that 
accompanies the excessive power causes 
the current to decrease. Some of the 
newer three-terminal regulators even 
have a rather “heroic," fail-safe mode 
built into them. They are designed such 
that should excessive power dissipation 
occur (which would cause destruction 
of the 1C) they fail as a short circuit. 
The result is a blown fuse farther back 
in the power supply. However, the 
circuit that is powered by the IC is 
never subjected to excessive, potentially 
destructive voltage. 

Since most of the design work is 
done by the manufacturer, our dis- 
cussion will deal mainly with practical 
applications of these components. The 
first consideration is to ensure that 
sufficient heat sinking is provided. The 
power dissipation will be determined by 
the current in the output and the 
voltage difference between the regulated 
output and the unregulated input. 
Another precaution that should be 
followed is proper bypassing. Under 
normal power supply construction this 
is of minimal significance. Only a O.I-/rF 
capacitor is required at the output. If, 
however, the regulator is to be located 
some distance from the unregulated 
supply, it is recommended that an elec- 
trolytic capacitor be placed across the 
input port. Usually, a value of 5 /jF is 
sufficient. 

Fig. 46A illustrates a 12-volt, 0.5-A 
regulated power supply which employs 
a National Semiconductor LM-341-12 
IC . U 1 should be affixed to a heat sink 
if heavy continuous currents are antici- 
pated. If only intermittent current loads 
are expected such as might be encoun- 
tered with a low power cw transmitter, 
the chassis will usually offer adequate 
heat sinking. Available also for the type 
of circuit shown are 3-A regulator ICs. 
They are contained in a TO-3 type of 
case. 

One virtue of most of the three- 
terminal regulators available is that very 
little current flows in the ground leg of 
die devices. Assume that an MC-7805 is 
available. This IC provides an output of 
5 volts, but is otherwise similar to the 
LM-341-12. This regulator could be 
employed in the 12-volt supply by using 
a resistive divider connected to the 
common pin of the IC. This variation is 
shown in Fig. 46B. In this application, 
die case of die MC-7805 must be 
insulated electrically from ground. 

When it is desirable to extend the 


160 Chapter 7 





Inside layout of the 500-mA supply. The pass transistor is seen on an L-shaped homemade heat 
sink. 



Fig. 49 - Schematic diagram of a 1 3-V, 0.5-A regulated supply. No overload protection is 
included, making it mandatory that the operator avoid dead shorts or heavy overloading at 
the output of the supply. T1 is rated at 1 A and has a 25-V secondary. CR1 through CR4 
are 50-PRV, 1-A diodes. VR1 is a 14-V, 1-W Zener diode. Q1 should be mounted on a large 
heat sink, at least 3X3 inches in size. 


current range of a regulated power 
supply beyond that of the regulator 1C. 
the circuit of Fig. 47 A can be used. In 
this example a series pass transistor. 01 . 
is “wrapped around" the IC to boost 
the current capability of the circuit. The 
operation of this circuit can be under- 
stood by noting the values of R1 and 
R2. Assume that the beta of Q1 is high. 
Most of the three-terminal regulator 
current will How through the I -ohm 
resistor and the diode, CR1. The offset 
voltage in CR1 is approximately the 
same as the emitter-base voltage of Ql. 
Because of this, the voltage drop across 
the 1-ohm resistor, R 1 , will be the same 
as that across R2. Since the ohmic value 
of R2 is 0.25 of R1 , four times as much 
current will (low in Ql as appears in the 
input terminal of Ul. The net result is 
that the current capability of the overall 
circuit is increased by a factor of 5. The 


i 


current limiting characteristics of the 1C 
are transferred directly to the composite 
circuit. 

Sometimes a power pnp transistor is 
not available in the home stock. Npn 
power transistors are much more com- 
mon. Fig. 47B shows a scheme for 
building a “synthetic” pnp power tran- 
sistor. This variation uses an npn power 
device with a smaller pnp transistor. 

A continuously variable 1.5-A regu- 
lated supply can be built as shown in 
Fig. 48. The LM317K IC can be used at 
any fixed output-voltage level by setting 
R1 to provide the desired output, V„ . 
Alternatively, R1 can be panel mounted 
to enable the builder to have a supply 
which can be varied from 1 2 to as 
much as 37 volts output. Ul of Fig. 48 
has built-in current and temperature 
limiting (thermal shutdown). A ripple 
rejection ratio of 80 dB is possible with 


this circuit. Output current limiting 
occurs at approximately 2.3 A. This 
much current could not be obtained at 
tlie higher output voltages. This is 
because of the relatively small value of 
filter capacitance used. The design rules 
for the unregulated power supplies 
which feed the regulators in Figs. 46 
through 48 are as given earlier in this 
section. 

A Low-Cost 1 3-V Supply 

Fig. 49 shows a practical circuit for a 
1 3-volt, 0.5-A regulated dc supply. It is 
housed in an aluminum Minibox, and 
some of the components are mounted 
on a homemade pc board in an effort to 
enhance compactness. 

There is no temperature compensa- 
tion or short-circuit protection circuitry 
included. The operator should exercise 
care by preventing crowbar conditions 
to exist at the power-supply output. 
Short-term overloads other than a dead 
short can be widistood for a few 
seconds without damage occurring to 
Ql, the pass transistor. Loads in excess 
of 500 mA will degrade regulation and 
cause excessive ripple in the output 
voltage. 

Output voltage amounts other than 
13 can be obtained by substituting 
suitable component values at R1 and 
VR1 (Fig. 12). Necessary information 
for the design changes was given earlier 
in this chapter. 

The 510-olun value listed for R1 was 
based on a minimum dc beta of 15 for 
Ql the value given in RCA’s data 
sheet for the 40251. The calculated 
value was 488 ohms, so the nearest, 
higher, standard resistance value was 
used. 510 ohms. The photographs show 
tlie general layout of the power supply. 
The container measured 3X4X5 
inches. The positive and negative termi- 
nals at the output are above chassis 
ground, thereby permitting the operator 
a choice of power-supply polarity. A 
third terminal is common to the case. It 
can be wired to the polarized terminal 
which will be employed as the common 



External view of the 12-V, 500-mA regulated 
dc supply. 

Test Equipment and Accessories 161 





Exterior view of the 1 2-V, 2-A regulated 
supply. 


bus for the equipment used with the 
regulated supply. 

A 2-A Regulated Power Supply 

Shown in Fig. 50 is a 2-A regulated 
dc power supply which can be adjusted 
to deliver 10 to 15 V. It is protected 
against overloads and short circuits. 
Output ripple is low, amounting to 10 
mV when a 2-A resistive load is con- 
nected across the output terminals. Reg- 
ulation and filtering remain good up to 
load conditions of 2.2 A. 

An 18-V, 3-A transformer was used 
for T1 in the example shown. It was 
obtained as a surplus item - brand and 
number unknown. However, it should 
be a simple matter to modify a 24-V 
transformer of suitable current rating, 
thereby obtaining an rms secondary 
voltage of 18. At the expense of overall 
efficiency, a 24-V transformer can be 
used with this circuit. 

The power supply is contained in a 
homemade aluminum case (see photo- 
graphs), which measures 3X5X6 
inches (HWD). A perforated top cover is 
used to permit the egress of heat from 
the transformer, regulator IC, and pass 
transistor. 

Q1 is mounted on a 3 X 4-inch heat 
sink. The latter is affixed to insulating 
hardware, as all three terminals of Q1 
must be above ground. 

R1 is adjusted for the desired dc 
output voltage. R2 is fashioned from 
No. 30 enameled wire. The required 
number of wire inches to provide 0.22 
ohm of resistance are scramble wound 
on a 10,000-ohm, 2-W resistor body. 
The resistor pigtails are used as termi- 
nals for the winding. U1 and some of 
the small components are installed on a 
homemade pc board. 

A Husky 12-V Power Supply 

Fig. 51 shows the schematic diagram 
of a 10-A regulated power supply which 
can deliver 11 to 14 volts of output. It 
was designed and built by W1GQO. 

Three 6.3-V, 10-A filament trans- 
formers are used with their primary 



Fig. 50 - Circuit details for a variable-voltage (10 to 15) 2-A regulated power supply which has 
overload protection. Resistors are 1 /2-W composition unless noted differently. CR1 through 
CR4 are 1 00-PR V, 6-A diodes. DS1 Is a 1 1 7-V neon lamp assembly. Q1 should be affixed to a 
large heat sink (3X4 inches or greater), and is a Motorola HEP248 or equivalent. R1 is a pc- 
mount control. R2 can be formed by winding a suitable amount of magnet wire on a short 
length of 1 /4-inch diameter insulating rod (see wire table in Handbook or ARRL electronics 
data book for wire resistance per foot). T1 should have an 18-V secondary with a 3-A or 
greater rating. A 24-V transformer can be used by removing a few secondary turns. Noise 
output is 10 mV under a 2-A load. U1 is a Motorola regulator IC. 



Interior view of the 2-A power supply. The pass-transistor heat sink is below the regulator 
board. 


windings in parallel. The secondaries are 
series-connected in the proper phase to 
provide a combined rms output of 18.9 
V. This causes a dc output potential 
from the bridge rectifier of 26.6 volts. 

There is nothing critical about the 
packaging format of this power supply. 
The important c chi side rati on is, how- 
ever, one of using heavy-gauge conduc- 
tors for point-to-point wiring in those 
circuits wlrich carry the full voltage and 
current of the unit: No. 14 or heavier 
hookup wire is recommended. The 
accompanying photograph shows how 
the power supply can be assembled to 


assure reasonable compactness. The case 
is homemade, and measures 3-3/4 X 6 X 
10 inches (HWD). 

02 is mourned on a home-built heal 
sink which was fashioned from I /32- 
inch thick aluminum. It measures 3 X 
1-1/2 inches. Similarly, the rectifier 
diodes (stud mount) are located on a 
homemade sink, 2X3 inches. Both 
handmade sinks have mounting feet 
formed by bending the stock at 90 
degrees to form an L bracket. The small 
part of each L is 3/4 inch deep. Q3, the 
main pass transistor, is placed on a 
finned heat sink purchased from Radio 


162 Chapter 7 





Fig. 51 - Schematic diagram of an 11- to 14-volt power supply with regulation, overload protection, and a 10-A rating. This circuit appeared 
first in OST for August, 1976, p. 26. CR1 through CR4are 50-PRV, 12-A diodes. Q1 is a 2N2905. Q2 is a 2N 3445. and Q3 is a 2N3772. U1 
is a National Semiconductor LM305 1C. T1 through T3 are 6.3-V. 10-A filament transformers connected so that the secondaries are in series 
(observe proper phasing). See text for data on R1 and R2. 


ie 

X 

it 

». 

X 

;r 

a 

h 

!t 

0 

II 

e 

a 

o 


Shack. It is 3-inches long and 2-inches 
wide. All of the heat sinks are bolted to 
the main chassis as an aid to heat 
transfer. Silicone grease is used between 
the sinks and the chassis, and between 
the transistor bodies and the heat sinks. 
Diodes CR1 through CR4 are treated in 
a similar manner. 

R1 is made by winding 9.7 feet of 
No. 22 enameled wire on the body of a 
IO-kf2, 2-W resistor. The desired output 
voltage is set by means of R2. The 
power supply has low ripple and is 
protected against overloads and short 
circuiting. 

Antenna Matching Techniques 

Most solid-state transmitting and re- 
ceiving equipment is designed to inter- 
face with a specific load impedance, 
respective to the antenna system. In 
most applications that impedance is 
between 50 and 75 ohms, assuming that 
unbalanced coaxial feed lines are used. 
Generally, coaxial feeders are used with 
single-band dipoles or gain types of 
antennas (beams). Multiband trap di- 
poles, beams, and verticals also dictate 
the use of coaxial feeders in most 
examples, although it is possible and 
practical to employ balanced two-wire 
feed systems with most of the antennas 
just mentioned. 


Because amateur transmitters and 
receivers are designed to operate at a 
particular antenna-impedance level, a 
matching network is used sometimes to 
effect maximum power transfer be- 
tween the antenna and the equipment - 
the purpose for creating a matched 
condition. Although many antennas can 


be matched at the feed point to the 
type of transmission line used, thereby 
eliminating the need for a matching 
network at the equipment end of the 
circuit, a matcher at the shack end of 
the system has some virtues. (1) A 
Transmatch (transmission-line matcher) 
enables the operator to maintain an 



View showing the interior of the 10-A regulated supply. 


Test Equipment and Accessories 163 












Fig. 52 - Examples of L and T types of 
matching networks. 


SWR of 1, or nearly so, over an entire 
amateur band without a need to re- 
adjust the match at the antenna feed 
point. Having a Transmatch at the 
equipment end of the circuit does not, 
of course, correct the mismatch at the 
antenna: It merely disguises the condi- 
tion so that the equipment sees the 
desired load impedance. (2) Depending 
upon tire kind of Transmatch used, 
harmonic energy from the transmitter 
can be attenuated by 30 dB or more as 
the signal passes through the matching 
system. This requires a low -pass or 
bandpass type of network. High-pass 
networks of the kind found in the 
Ultimate Transmatch, popularized by 
W1 ICP in QST for July, 1970, are of 
less value in this regard, despite the wide 
range of impedances they are capable of 
matching. Fig. 52C shows the basic 


high-pass T network. At D is the modi- 
fied configuration described by W1ICP. 

Shown also in Fig. 52 are two forms 
of L network which are useful in match- 
ing the equipment to a transmission 
line. All of the equations shown in Fig. 
52 are based on matching loads to 
sources which are, respectively, pure 
resistances. The L and C components 
for the circuits are illustrated as being 
variable. In a practical situation the load 
presented by the transmission line is 
purely resistive at only that frequency 
in the amateur band for which the 
antenna is constructed and matched to 
its feeder. Therefore, as the operating 
frequency is moved above or below that 
at which an SWR of 1 exists, the load 
becomes reactive. Should the reactance 
become great enough in magnitude to 
result in a high SWR, say, 2:1 or greater, 
the transmitter may not load into the 
antenna system effectively, thereby en- 
dangering the output transistors (if SWR 
protection is not included in the PA 
stage). A high SWR will also reduce the 
power transfer to the load. Similarly, if 
the receiver front end has a filter which 
was designed for the characteristic im- 
pedance of the transmission line (usually 
50 ohms), the mismatch will degrade 
the filter performance. Because of the 
foregoing considerations it is necessary 
to make the L and C elements of the 
network variable to permit matching to 
loads which exhibit unknown reac- 
tances. These reactances are reflected to 
the equipment end of the feed line by 
the antenna when a mismatch is present 
at tiie feed point. As the mismatch at 
the antenna increases so does the loss in 
the feeder. The higher the operating 
frequency, the more pronounced the 
loss condition becomes. In situations 
where a high SWR must be accepted, as 
may be the case in some portable or 
emergency operations, high-quality 
(low-loss) feed line should be used. If 
the feeder length is less than 50 feet at 
frequencies in the hf and mf spectrum, 
RG-58/U and RG-59/U should be suit- 
able with respect to losses versus SWR. 
Subminiature coaxial cable (RG-174/U 
type) is not recommended except when 
other types of cable are too heavy. For 
feed-line runs greater than approxi- 
mately 50 feet, RG-8/U or RG-ll/U 
cable is a better choice, even when the 
SWR is not high. Open-wire feeders will 
have the lowest loss factor of the 
numerous kinds of transmission lines 
because the dielectric material is air, 
principally. Feeder losses and impedance 
matching are especially significant when 
QRP equipment is being used every dB 
counts! 

The T networks at C and D of Fig. 
52 are capable of accommodating a 
much greater range of jmpedances than 
would be possible with L or pi net- 
works. For field work this is an impor- 



Fig. 53 — Bandpass types of matching net- 
works. These are used frequently in Trans- 
matches. They offer harmonic rejection. 


tant consideration, for makeshift an- 
tennas are often used during portable 
operations. The equations given are 
based on a loaded Q of 5, which is an 
arbitrary figure picked by the writers. 
Other values of Q would be acceptable, 
but the low figure of 5 has proved to be 
practical in the interest of matching- 
network bandwidth. More specifically, 
the higher Qs require that the Trans- 
match be readjusted even when small 
changes in operating frequency are 
made. The higher the network Q, the 
more critical the adjustment procedure 
- another consideration. A Q of 5 is a 
practical ball-park figure, and yields 
practical L and C values for a wide range 
of impedance conditions. 

T-network Transmatches of the type 
shown have the advantage of rejecting 
frequencies below the one to which 
they are tuned. Therefore, the high -pass 
characteristic can be used to advantage 
in rejecting bc-band energy which could 
affect the performance of a receiver. 
Those who live near be stations often 
experience problems with receiver over- 
loading and IMDwhen operating on 160 
or 80 meters. 

Other Matching Networks 

Operators who wish to take ad- 
vantage of the harmonic-suppression 
characteristics of a bandpass type of 
Transmatch may elect to use one of the 
circuits shown in Fig. 53. A bandpass 


164 Chapter 7 





if tapped every 4 turns (3022 Miniductor). L3 Is a toroid inductor with 35 turns of No. 20 
enam. wire on an Amidon T1 30-2 core. SI is a single-pole. 10-position rotary ceramic wafer 
switch with the shaft and collar insulated from ground. Z1 is the circuit of Fig. 15. 


network will also aid reception through 
rejection of frequencies above and be- 
low the one to which the network is 
tuned. 

At Fig. 53A is an unbalanced band- 
pass matching network that can be used 
between the station equipment and the 
coaxial feeder. Alternatively, it can be 
placed between a single-wire antenna 
(resonant or random length) and a 
coaxial feed cable to the amateur 
station. Reactance values are given to 
permit calculation of the /, and C values 
for a given band of operation. For 
multiband use, C and I, should be 
chosen for the lowest operating fre- 
quency anticipated. In such an event, 
taps should be placed on LI to permit 
matching at the high end of the Trans- 
match frequency range. LI and Cl must 
be able to form a resonant circuit it the 
operating frequency. Likewise with L2 
and C2. The tap on L2 is moved 
experimentally, along with adjustment 
of Cl and C2, to obtain an SWR of 1 . 

The operating principle and adjust- 
ment procedures are the same for the 
circuit of Fig. 53B. In this example the 
Transmatch is designed to accommodate 
balanced feeders, such as would be used 
with an end- or center-fed Zepp an- 
tenna. The ARRI. Antenna Book con- 
tains in-depth descriptions of various 
antennas that can be used with these 
Transmatch circuits. 

For multiband use of the network in 
Fig. 53B, it will be necessary to tap Cl 
toward the center of L2 as the operating 
frequency is increased. Similarly, taps 


should be placed on LI . Respective to 
all of the matching circuits shown here, 
the wire size of the inductors and the 
plate spacing of the variable capacitors 
must be adequate for the power level 
employed. The wire size should be great 
enough to minimize IR losses and heat- 
ing. Capacitor plate spacing should be 
such that arcing does not occur during 
periods of high SWR — as encountered 
during system adjustment. 

Transmatch Adjustment 

Precise adjustment of a Transmatch 
is done best by applying transmitter 
power and observing an SWR indicator 
while adjusting the network. Tuning 
should be done at the lowest power- 
output level practicable, thereby mini- 
mizing damage to the PA stage and 
lessening the chance of causing QRM to 
those who may be using the frequency. 
Various kinds of SWR indicators are 
suitable for use with Transmatches, but 
for on-the-nose adjustments the instru- 
ment should have high sensitivity: Full- 
scale deflection of the indicating meter 
should be possible at the low-power 
level used during initial setup of the 
Transmatch. In this regard the circuit 
treated by Bruene in QST for April, 
1959 is excellent. He described the 
design features of a directional watt- 
meter which used a toroidal current- 
sampling transformer in an rf bridge 
circuit. Practical examples of that type 
of instrument were given earlier in this 
chapter and in QST for December, 
1969. Circuits were described for power 


levels from 5 to 1000 watts. There are 
two distinct advantages offered by the 
Bruene circuit over that of the so-called 
Monimalch SWR meter described by 
McCoy in the 1950s (QST). The latter 
exhibits extreme frequency sensitivity, 
with declining sensitivity as the opera- 
ting frequency is lowered. Instruments 
of that kind are not suitable for QRP 
work unless a meter amplifier is used. 
Additionally, it is difficult to employ 
the Monimatch circuit as a calibrated 
wattmeter because of its frequency 
sensitivity. The Bruene circuit, however, 
is suited to the purpose in an ideal 
manner. An SWR indicator of this 
variety can be used for Transmatch 
adjustment and for measuring rf power. 
Fig. 15 shows a practical circuit for a 
10- to 1000-W version of the bridge. 
Fig. 17 shows the schematic diagram of 
another version of the instrument - 
capable of full-scale deflection at 1 
watt. Fach of the examples are suitable 
for use when adjusting Transmatches. 

In a practical situation, the SWR 
indicator is placed between the trans- 
mitter and the Transmatch. The indica- 
tor is set for maximum sensitivity in the 
refiected-powcr position. Transmitter 
power is advanced to obtain a few 
divisions of meter deflection. The Trans- 
match controls are adjusted to cause a 
meter reading of zero. The transmitter is 
retuned for maximum PA output with- 
out increasing the drive. Next, the SWR 
indicator is set for a forward-power 
reading and the sensitivity control is 
adjusted for a full-scale meter reading. 
Then, the operator returns the bridge to 
the reflected-power mode and makes 
final adjustments with the Transmatch 
to secure zero meter deflection. Normal 
operating power can be established now, 
setting die sensitivity control of the 
bridge for full-scale indication on the 
meter (forward-power mode). Bridges 
which are intended for rf-power reading 
do not necessarily have sensitivity con- 
trols on the instrument panel. There- 


. 



Fig. 55 - Exterior view of the Transmatch as 
seen in its homemade aluminum case. The 
control at the upper right is not used. 

Test Equipment and Accessories 165 



fore, adjustments of the Transmatch 
must be made while utilizing whatever 
amount of meter-scale deflection is 
available. 

In situations where the meter will 
not drop to zero, no matter how care- 
fully the Transmatch is adjusted, it will 
be likely that the transmitter is putting 
out considerable harmonic energy. Even 
though a perfect match has been 
effected at the desired operating fre- 
quency, the harmonic energy is being 
reflected back to the bridge, causing a 
false indication that high SWR exists. A 
remaining cause of imperfect meter 
zeroing can be brought about by a 
bridge that was not nulled properly at 
the operating frequency. That is, al- 
though it had a characteristic impedance 
of 50 ohms at some frequencies in the 
hf spectrum, internal unwanted reac- 
tances in the bridge circuit could make 
the instrument other than 50 ohms at 
some specified frequency. The effect is 
one of not getting a reading of zero 
when an SWR of 1 exists in a 50-ohm 
feeder system. 

Fig. 54 shows the circuit of a modi- 
fied T-network Transmatch of the kind 
illustrated in Fig. 52D. It is designed to 
operate from 80 through 10 meters at 
power levels up to 150 watts contin- 
uous. Although Cl is a dual-section 
capacitor assembly, configured as a 
dual-differential variable, a single capac- 
itor can be used to form the circuit of 
Fig. 52C. The dual-differential capacitor 
arrangement of this circuit was em- 
ployed for experimental purposes. In 
practice there is little difference in the 
matching ranges of the three circuits. A 
rotary inductor can be used in place of 
the tapped coil and switch shown, and 
will ensure a greater impedance- 
matching range than the tapped coil 
will. Transmatches of tliis type should 
always be adjusted so that the maxi- 



Fig. 56 - Interior of the Transmatch. Cl A 
and Cl B are joined by means of a right-angle 
drive. Insulated shaft couplings are used at SI, 
Cl and C2 (Millen 39016). SI is mounted on 
a phenolic plate (center of picture). An unused 
coaxial connector is visible at the lower center. 
Z1 is at the upper right. 


mum practical amount of C is used at 
C2 during a matched condition. The 
tighter coupling will provide greater 
Transmatch efficiency (lower insertion 
loss), and will lower the circuit Q by 
virtue of tighter coupling to the load. 
The latter will lessen the need of re- 
adjusting the Transmatch when small 
changes in operating frequency are 
made. It should be noted that an SWR 
of 1 can be obtained at various settings 
of the controls, but always use as much 
capacitance at C2 as is possible, consis- 
tent with an SWR of 1 . Figs. 55 and 56 
show how the Transmatch is built. A 
Bruene type of rf bridge is included in 
the box to permit monitoring of the 
SWR. The assembled unit measures 
(HWD) 4-1/2 X 8 X 7 inches, and has a 
homemade aluminum cabinet. 

Fig. 57 A illustrates a QRP Trans- 
match which is suitable for power levels 
up to 25 watts. Because of its small size 
it is ideal for field applications. An 
external SWR indicator is needed with 
this unit. A homemade variable induc- 
tor, designed and built by K1KLO, is 
die heart of the matcher. It contains 
one half of a powdered-iron toroid core 
(1-inch diameter core, No. 2 iron mix, 
wall height and thickness of 3/ 16 inch). 
The core material moves in and out of a 
hand-wound coil which contains 32 
turns of No. 22 enamel wire, 7/1 6-inch 
OD. A detailed description of this 
Transmatch was published in QST for 
February, 1976. Fig 57B shows a 
method for adding 80-meter coverage. 

A slug-tuned coil (LI of Fig. 57 A) is 
switched in parallel with the half-toroid 
one (L2 ) to lower the inductance during 
operation on 20, 15, and 10 meters. The 
former has an inductance range of 3 to 
9 /ill. The slug-tuned inductor has a 3.1 - 
to 4.8-/dl range. 

Simplification of the circuit will 
result if Cl is replaced by a single 
365-pF unit of the type used at C2 . The 
resulting circuit would be similar to that 
of Fig. 52C. This Transmatch is housed 
in a 1-1/2 X 2-3/4 X 4-inch plastic 
meter case. Phono connectors are used 
for the input, output and ground ter- 
minals. Alligator clips have been sol- 
dered to phono plugs to facilitate con- 
nections to earth ground and a single- 
wire antenna, if the latter is used. Figs. 
58 and 59 show how the unit is built. 

A 40-Meter Transmatch 

Fig. 60 shows the circuit of a QRP 
Transmatch for use on 40 meters. The 
input circuit is arranged for switching a 
resistive bridge in series with the 
matching network during adjustment 
for an SWR of 1. Cl, C2 and LI 
comprise a high-pass network for 
matching a wide range of impedances to 
a 50-ohm source. 

During normal operation SI is 
placed in the operate mode, bypassing 



Fig. 57 — The diagram at A is for the 40- 
through 10-meter Transmatch. At B, a 
suggested circuit for coverage from 80 
through 10 meters. 

Cl — Dual-section air variable (Miller 2109, 

J. W. Miller Co.. 19070 Reyes Ave., 
Compton. CA 90224). See text. 

C2 — Calectro or Archer single-section minia- 
ture 365-pF variable. 

J1 -J3, incl. — Phono jack. 

LI — 3.1- to 4.8 -mH slug-tuned inductor 
(Miller 4504 with red core). 

L2 — See text. Contains 32 turns of no. 22 
enam. wire, air wound, 7/16-inch OD. 

L3 - 5.5- to 8.6-pH slug-tuned inductor 
(Miller 4504 with red core). 

SI , S2 — Spdt slide or toggle switch. 



Fig. 58 - Exterior view of the QRP 
Transmatch. J1 , J2 and J3 are seen at the 
far right. 


166 Chapter 7 



1 



Fig. 59 — Interior view of the Transmatch 
showing the K1 KLO variable inductor at 
the lower center. 


the bridge. Any meter with a sensitivity 
of 50 to 500 n A will be suitable at M 1 . 
The instrument used in this example 
was borrowed from a junked tape 
recorder. CR1 is a germanium diode of 
the IN34A variety. 

J3, a single-terminal binding post, is 
connected in parallel with coax connec- 
tor J2 to permit attachment of a single- 
wire antenna. The assembled unit is 
contained in a small aluminum chassis 
(5X3X1 inches). A smaller case can 
be used if a more compact assembly is 
desired. Fig. 61 shows how the compo- 
nents are arranged in the box. 

Assorted Test Equipment 

This section contains a collection of 
circuits that have been built by the 
writers for their own use. Many of the 


details are not presented, because they 
will depend upon the characteristics of 
the parts used by the builder. The junk 
box and surplus market can provide 
many of the needed components. 

Noise Generator 

Shown in Fig. 62 is a circuit for a 
noise generator. This unit was inspired 
by an investigation of the effects of 
Zener doides on the noise performance 
of amplifiers. The experiments sug- 
gested that Zener diodes were not opti- 
mum for biasing very low-noise ampli- 
fiers. This was due primarily to noise- 
modulation effects when strong signals 
were present, rather than actual degra- 
dation of noise figure. 

There is an expression among design 
engineers when a problem is encoun- 
tered. “If you can’t lick the problem, 
feature it.’’ This was the policy that was 
followed in the noise generator shown. 
The major noise source is CR1, a 5.1- 
volt Zener diode that is used to bias the 
first amplifier. Since no bypassing of the 
Zener is used at the base, and the 
current in the diode is small, the exces- 
sive noise currents in the diode will flow 
through the base of the transistor. The 
amplified output is applied to a second 
stage of gain. The second amplifier has a 
51 -ohm resistor in the collector to 
provide a controlled output impedance. 

The noise output of this circuit has 
been measured on a spectrum analyzer. 
The detailed distribution of noise with 
frequency will not be presented since it 
will vary considerably with Zener diode 
and transistor characteristics. Generally, 
the noise in the hf region was quite 


it . ii 1 At 



Fig. 61 — Interior view of the 40-meter 
Transmatch. 


robust, reaching levels of 80 dB higher 
than the noise output from a room- 
temperature resistor. The noise output 
is still 20 dB above a 290° K resistor at 
432 MHz. 

The builder should not attempt to 
estimate noise figure with a device as 
crude as this one. It may be used, 
however, as a source for tuning receivers 
or amplifiers. If one were to build a 
free-running multivibrator, using a 555 
timer, with a total period of 1 to 2 
seconds, it could be used to auto- 
matically turn the generator on and off. 
The system could then be used in 
conjunction with a step attenuator to 
adjust a vhf preamplifier for low noise 
figure. The output detector would be 
the operator’s ears, although refined 
circuitry could be built for the purpose. 

Audio Voltmeter 

Shown in Fig. 63 is a circuit for an 
uncalibrated audio voltmeter. Two 741 
operational amplifiers are used. The first 
one is an amplifier with a voltage gain of 
11. The output of this stage has a pair 
of attenuators that may be switched 
into the system. The second amplifier 
contains a meter within a bridge recti- 
fier. Since the rectifier is in the feed- 
back loop of the op amp, diode charac- 
teristics are not critical. The diodes 
should all be of the same type, though. 

Calibration of the attenuators is 
straightforward, although unusual for 
audio applications. First, a 50-ohm resis- 
tor is placed temporarily across the 
input. Then, an audio generator is ob- 
tained and set for a sine-wave output of 
several volts. A 50-ohm resistor is placed 
in series with the audio generator out- 
put, if the output impedance is as low as 
would be the case with an op amp 
output. Then, a 50-ohm step attenuator 
is set for 30 dB of attenuation and 
placed between the two units. Power is 
applied to each, and the input control is 
set for a full-scale meter reading. The 
attenuator controlled by SI is set to die 
-3 dB position, and the 50-ohm step 



Fig. 60 — Schematic diagram of the 40-meter Transmatch. Resistors are 1/2 -W composition. LI 
contains 30 turns of No. 22 enam. wire on a T68-2 toroid core. SI is a dpdt slide switch. 


Test Equipment and Accessories 167 






Fig. 62 - Circuit details of a noise generator. Fig. 63 - Details of the audio voltmeter. CR1 through CR4 are 1N914s. Ml is a 500-/iA meter, 
Excess-noise output is greater than 70 dB at and SI is a center-off spst toggle switch. 

14 MHz, and is detectable at 432 MHz. 


attenuator is adjusted to increase the 
power to the audio voltmeter by 3 dB. 
The 5000-ohm control is then adjusted 
for a full-scale reading on the meter. 
The same procedure is used for calibra- 
ting the — lOdB position of SI . 

Typically, this meter is used for 
receiver testing. It can be used without 
the attenuators for alignment. If a 
measurement of MDS is to be per- 
formed, the gain in the receiver and the 
level control in the voltmeter are set for 
a full-scale reading from the noise out- 
put of the receiver. Then, a signal 
generator is applied to the input of the 
receiver, and SI is thrown to the -3 dB 
position. The rf signal generator is set to 
again yield a full-scale reading. The 
output power available from the rf 
generator is then the MDS of the receiv- 
er. The signal required to obtain a 10-dB 
signal plus noise -to-noise ratio can be 
evaluated in a similar way by using SI in 
the -10 dB position. 

One refinement that the builder 
should consider is to add a 47-kS2 
resistor between the output of U1 and 



A general-purpose test instrument. The meter 
at the left is for readout of an audio volt- 
meter. The instrument is not calibrated, but 
there are calibrated attenuations of 3 and 1 0 
dB. The meter at the right is an indicator for 
a broadband rf detector. A broadband ampli- 
fier is contained in the case. It allows sensi- 
tivities of -65 dBm up through 50 MHz. The 
rf detector is used in conjunction with a step 
attenuator to produce 1-dB accurate measure- 
ments of gain, return loss and related param- 
eters. 


the electrolytic capacitor connected to 
SI . This will keep the dc potential on 
the capacitor equal to that at the output 
of U1 , preventinga large transient when 
SI is thrown into one of the attenuation 
positions. 

Capacitance Bridge 

A simple capacitance bridge is shown 
in Fig. 64A. This unit is useful for 
determining the value of unmarked 
capacitors, such as those of the “dog 
bone” ceramic variety. The audio source 
may be from an audio oscillator, a 
square-wave oscillator, or even the sta- 
tion receiver which is tuned to a steady 


carrier. The audio input signal is applied 
to a transformer, Tl. The secondary of 
this transformer is allowed float with 
respect to ground. However a 50-kft 
control is placed across the secondary 
with the arm attached to ground. The 
“unknown” capacitor is- placed in series 
with a capacitor of known value to form 
a bridge configuration. This is empha- 
sized in the circuit of Fig. 64B. 

The junction of the two capacitors is 
connected to a high input inpedance 
JFET audio amplifier. Nearly any avail- 
able FET should be suitable for this 
application. In use, the control is tuned 
until minimum output is noted in the 



Fig. 64 - Schematic diagram of the capacitance bridge. 


168 Chapter 7 






Fig. 65 - Circuit of the weak-signal 14-MHz generator. LI has 24 turns of No. 26 enam. wire on 

an Amidon T50-6 toroid core. The link consists of a single turn of wire. 


headphones. The depth of the null is 
quite large in our unit. 

In order for this bridge to be useful, 
il is necessary that the 50-kf2 control be 
linear and calibrated. In our unit, a 
10-turn control is used with a turns- 
countingdial. If the output of the dial is 
interpreted as a ratio between 0 and 1 , 
the unknown capacitor is related to the 
standard capacitor with the equations 
shown in Fig. 64. If the two capacitors 
are equal, the bridge will be balanced 
when the control is set at midrange, 
where R = 0.5. If the builder does not 
hdVe a 10-turn control with a turns- 
counting dial, a more mundane system 
could be calibrated with a handful of 
capacitors of known value. 


The bridge will operate over a wide 
range of capacitance. Using a 10-pF 
standard, very small values are easily 
determined. An example would be the 
parallel capacitance of a quartz crystal. 
Values of up to 0.1 -juF have been 
measured as well. The best accuracy will 
always be obtained when the standard 
capacitor is close in value to the capac- 
itor being measured. A group of 
known-value capacitors with 1 -percent 
tolerance are kept on hand. Three 
binding posts are provided on the instru- 
ment for easy insertion. 

Low-Level RF Source 

When working on receivers, one of 
the most useful pieces of test gear one 


can use is a low-level signal source. 
While there are signal generators avail- 
able that will do the job nicety, they are 
expensive. The less expensive kit genera- 
tors are not too suitable for precise 
receiver work, since they have too much 
leakage to allow the measurement of 
weak signals. If one ever has the chance 
to observe the level of shielding and 
decoupling that is used in a high-quality 
signal generator, he will realize why 
inexpensive generators are so leaky. 

All is not lost meaningful weak- 
signal measurements can be made in the 
home shop. Shown in Fig. 65 is the 
circuit of a 14-MHz source. The key to 
good performance is the shielding. The 
generator is built in a box made from 
double-sided pc-board material. A high 
quality feedthrough capacitor is used to 
get power into the box, and thorough 
power supply decoupling is applied 
within the unit. Extensive attenuation is 
used within the oscillator housing with 
shield partitions between the sections of 
the attenuator. A battery is used to 
power the unit, thereby avoiding signals 
that could leak along signal ground 
paths in the power lines. 

The best way to calibrate this source 
is by using a better generator in con- 
junction with a receiver. The age in the 
receiver is defeated and an audio volt- 
meter is used to monitor the receiver 
output. The resistors in the attenuator 
are picked to provide an output that 
corresponds to a reasonably weak signal 
(S5 or thereabouts). The box is soldered 
shut with the crystal inside the shielded 
enclosure. The level in the receiver is 
carefully noted on the audio voltmeter. 
Then, the signal generator is substituted 
in place of the crystal-controlled source, 
and is adjusted for an identical output 
response. The output is noted, then 
marked on the outside of the box. 

This source is now usable in the shop 
in conjunction with step attenuators for 
the measurement of receiver MDS. We 
have been able to duplicate laboratory 
results within 1 dB with these methods. 
It should be mentioned that even if 
calibration is not possible, a source of 
the type described can be useful for 
comparative measurements. Further- 
more, since the calibration may be done 
with a generator that might be too leaky 
to be useful at really low levels, die 
techniques may be applied to extend 
the measurement capabilities of a 
moderately equipped home shop. 



Exterior of the signal generator. It provides 
low-level output for 7 and 14 MHz. 



Test Equipment and Accessories 169 






I .uu> 



S M -SILVER MICA 


Fig. 67 — Circuit details for a wide-range rf 
oscillator (see text). 


It is not necessary that the units be 
confined to a single band. One source 
was built which used a 7 -MHz crystal in 
the circuit shown, but had the tuned 


circuit peaked at 14 MHz. Both of the 
outputs were calibrated, resulting in a 
two-band source. 

Crystal-Controlled Sources for 
IMD Measurements 

In the evaluation of the two-tone 
dynamic range of a receiver, the two 
parameters needed arc the input inter- 
cept and the MDS. The MDS can be 
measured with the weak-signal source 
just described, and a step attenuator. 
For evaluation of the input intercept, or 
for evaluating the dynamic range 
directly, and then calculating the equiv- 
alent input intercept, a pair of stronger 
sources are needed. The frequencies 
should be separated by 20 kHz. 

A suitable circuit for such sources is 
shown in Fig. 66. This circuit should be 
well shielded, although the requirements 
are certainly not as severe as with the 
weak-signal source. Of greater signifi- 
cance is that tlie sources be well 
decoupled from the power supply lines 
and that the buffering be effective. In 


the source shown, output buffering is 
achieved with a cascode amplifier. This 
circuit was chosen because of the low 
feedback capacitance. Because of this, 
the impedance seen at the input of the 
buffer is virtually independent of the 
load or signals present at the output. A 
dual-gate MOSFET would probably do 
an excellent job as an output buffer as 
well, and is certainly capable of deliv- 
ering 10 mW of output power. In tire 
circuit shown in Fig. 66, Rl is picked 
for an output power of +10 dBm. A 
low-pass filter is used in the output to 
ensure that the power measurements 
indicate the power available at 14 MHz 
and not be influenced by harmonic 
content. Also, harmonics could, in some 
cases, confuse the IMD results. 

The nature of the measurements 
were described in chapter 6 in connec- 
tion with our discussion of dynamic 
range and the intercept concept. Two of 
the generators of the type shown in Fig. 
66 are required, and with equal output 
powers. The two outputs are added in a 



Fig. 68 — An elaborate version of the circuit shown in Fig. 67. 

170 Chapter 7 









Outside view of the wide-range test oscillator. 


6-dB hybrid combiner (described in this 
chapter as a return-loss bridge). The 
output of the “hybrid” is applied to a 
50-ohm step attenuator and then to the 
receiver being tested. The reason for 
using die “hybrid” and the extremes of 
buffering is to prevent one generator 
from being phase modulated by the 
other. This effect is detected easily as a 
difference in the IMD levels at die two 
distortion frequencies. A good precau- 
tion (besides those oudined) would be 
to use separate battery packs for power 
of each of the generators. 

Tunable RF Generators 

For many measurements a tunable 
source of rf is desired. Applications 


would range from antenna evaluation 
and impedance measurement with a 
return-loss bridge, to measurement of 
die resonant frequencies of tuned cir- 
cuits. Shown in Fig. 67 is an FET 
oscillator that is capable of operation 
over a wide range of frequencies. The 
Colpitts configuration is used widi a 
split stator variable capacitor. With most 
capacitors used for tuning, a frequency 
range of over 3:1 may be covered with a 
single toroid coil. If the capacitor has a 
reasonable low minimum capacitance 
(10 pF or so, including strays) the 
oscillator will operate at frequencies up 
to about 250 MHz. Toroidal coils or 
air-wound inductors may be used. A 
6-volt lantern battery is suitable for 
power. 

This oscillator may be used for 
evaluating luned circuits by placing a 6- 
or 10-dB attenuator at the output. The 
output of the pad is then applied to a 
link on the unknown resonator. A sensi- 
tive rf detector is loosely coupled to die 
resonator, and the oscillator is tuned for 
a peak response. 

Band-switching versions of this oscil- 
lator may be built. However, it is 
important that all three of the hot leads 
of coils be switched. If they are not the 
stray resonances in the larger coils used 
for the lower frequencies may cause the 
output level to vary on the higher 
ranges. 

A more elaborate oscillator is shown 
in Fig. 68. This unit is band-switched to 
cover a range of 1.7 to 15 MHz, hitting 
the four lowest amateur bands. Q1 is a 
FET that serves as a simple Hartley 
oscillator. It is normally tuned over the 
range of 3 to 6.5 MHz, using a single 
section of a BC455 surplus receiver 
capacitor. The oscillator is moved to 



Fig. 69 - The circuit at A is the field-strength meter. LI has 20 turns of No. 26 enam. wire on 
an Amidon T50-6 toroid core. The tap is located 5 turns above ground. Cl is a subminiature 
transistor-radio type of variable. At B is the 1 00-kHz standard. C2 is a 45-pF mica or ceramic 
trimmer. Y1 is an International Crystal Mfg. Co. type GP crystal. 



Fig. 70 — Exterior view of the test unit. A 
small Minibox serves as a case. 


lower frequencies by switching in the 
parallel combination of the other two 
capacitor sections. To move the oscil- 
lator to liigher frequencies, additional 
inductors are paralleled with the main 
one. 

The gear-drive mechanism built into 
the surplus tuning capacitor provides 
more than adequate bandspread. How- 
ever, for special situations, even finer 
tuning is desirable. Tliis is realized with 
a back-to-back pair of varactor diodes. 
In the circuit shown, a Motorola MV 104 
dual is used, with both diodes in the 
same package. The diodes are tapped 
well down on the tuned circuit in order 
to provide high tuning resolution. The 
varactors may be controlled front one of 
two separate sources, which are selected 
by a switch. One is a 10-turn 50-kf2 
control that is biased with the 6-volt 
regulated supply used for the oscillator. 
The other is a swept voltage source 
consisting of a large electrolytic capaci- 
tor. a charging resistor to the 12 -volt 
supply, and a push-button to initiate the 
sweep. The tuning range of the Varicaps 
is very restricted, covering only about 7 
kHz on the 80-meter band. The main 
application for this absurd level of 
resolution was for the evaluation of 
homemade crystal filters. 

Output buffering is handled with a 
two-stage amplifier. Q2 serves as a 
source follower to drive Q3 which is a 
fed-back power stage. A separate atten- 
uated output is provided on the panel of 
the generator to drive a frequency coun- 
ter. 

Exact component values are not 
given for the tuned circuits. They will 
depend upon the parts the builder has 
on hand. All of the coils are wound on 
Amidon toroids. The main resonator is 
wound on a T68-2, with T50-6 cores 
being used for the high-frequency coils. 
The tap on the main resonator coil 

Test Equipment and Accessories 171 




should be about 0.25 up from the 
ground end. 

A Handy Field Tester 

The matter of including a 100-kllz 
secondary frequency standard in each 
receiver built can be costly. A good 
alternative is to have a separate assem- 
bly that can be used with any receiver, 
thereby reducing the cost which would 
result from purchasing several crystals. 
Fig. 69 B shows a 100-kHz FET oscilla- 
tor which operates from 9 volts. A short 
length of wire can be connected be- 
tween J2 and the input terminal of the 
station receiver to provide 100-kHz 
markers. C2, a 45-pF mica trimmer, is 
used to zero beat the oscillator with 
WWV. 

Contained in the same 1-1/4 X 2-1/4 
X 4-1/4-inch Minibox is the circuit of 
Fig. 69 A. It is a tunable field-strength 
meter with a range of 7 to 29 MHz. No 
provisions have been included for cali- 
bration of the instrument. It functions 
only as a relative-indicating meter, but is 
useful in the field for “sniffing” rf in 
equipment, and for determining if an- 
tennas are functioning properly. It also 
enables the user to get a reasonable idea 
of what a near-field antenna pattern 
looks like. This assembly was built 
especially for QRP DXpeditions, where 
lightweight test gear is desirable. 

LI in Fig. 69 A consists of 19 turns 
of No. 24 enameled wire on an Amidon 
T50-6 toroid core. The diode tape is 
placed 5 turns up from the ground end 
of the coil. This prevents the rectifier 
diode from loading the tuned circuit. A 
short piece of hookup wire, or a whip 
made from brazing rod or piano wire, is 
inserted into J1 for sampling rf. Cl is 
tuned for a peak response at the oper- 
ating frequency, as indicated at Ml . Cl 
is a miniature 365-pF variable of the 
variety used in transistorized a-m band 
receivers. 

Y1 of Fig. 69B is an International 
Crystal Co. unit of the general-purpose 
type. Load capacitance is 30 pF. Any 
crystal with similar characteristics 
should work satisfactorily in the circuit. 
Fig. 70 shows an exterior view of the 
assembled tester. 

Transistor and Crystal Testers 

Fig. 71 contains the circuit of a 
“go-no-go” type of transistor tester 
winch can be used to determine whether 
transistors are defective, npn or pnp 
varieties, or FETs. A fundamental type 
of crystal is used at 20 MHz to permit 
the devices under test to function as 
oscillators. Output from the oscillator is 
rectified by a voltage doubler (CR1 and 
CR2). The dc voltage is routed to a 
50-nA meter, Ml, to provide a visual 
indication of performance. S3 is used to 
apply forward bias to bipolar transis- 
tors. It is switched to the open position 



Fig. 71 — Diagram of the FET and bipolar-transistor tester. Resistors are 1 /4- or 1 /2-W com- 


position. Capacitors are disk ceramic. 

BT1 — Small 9-V transistor radio battery. 
CR1, CR2 — 1N34A germanium diode or 
equiv. 

J1 - Four-terminal transistor socket. 

J2, J3 — Three-terminal transistor socket. 

Ml — Microampere meter. Catectro D1 910 
or similar. 

R5 — 25,000-ohm linear-taper composition 


control with switch. 

RFC1 - 2.5-mH rf choke. 

51 — Two-pole double-throw miniature 
toggle. 

52 — Part of R5. 

53 — Spst miniature toggle. 

Y1 — surplus crystal. 


for testing FETs. SI reverses the battery 
polarity for testing npn or pnp transis- 
tors. At the voltage levels available in 
the tester, damage will not occur to any 
transistor, regardless of the positions of 
SI and S3. 

Three different styles of transistor 
socket are placed on the top panel of 
the tester ( J 1 , J2 and J3) to accommo- 
date the three most popular lead 
arrangements. TP1 is available for scope 
attachment, should the user wish to 
measure the output voltages of a group 
of similar transistors. This will give a 
general idea of the gain comparison 
between units the higher pk-pk levels 
indicating greater small-signal gain. 

This tester is useful only for testing 
transistors whose f T characteristics are 
50 MHz or higher. Although most tran- 
sistors will function as oscillators at 
some frequency lower than the rated f T , 
the test results with the circuit of Fig. 
7 1 will not be of value. 


The tester illustrated schematically 
in Fig. 72 will help the user to deter- 
mine the relative quality of crystals. It is 
set up as a Pierce oscillator, and three 
fixed-value feedback capacitors can be 
selected by means of SI. The feedback 
capacitor chosen will depend on the 
frequency of the crystal under test, and 
on its activity characteristic. 

Visual readout is handled in the 
manner described for the circuit of Fig. 
7 1 . TP1 can be used for connection to a 
scope, or a short antenna can be 
attached to the test point to permit use 
of the tester as a frequency marker. 

Overtone crystals can be checked in 
this unit, but they will oscillate at their 
fundamental modes. A polarity rever- 
sing switch, S3, permits use of npn or 
pnp transistors at Q1 . A transistor 
socket is located on the top panel of the 
tester, thereby making the tester useful 
for checking transistors of unknown 
characteristics. J1 through J4, inclusive, 


172 Chapter 7 



i 



Fig. 73 - Photograph of the two testers. They are housed in homemade aluminum cases. The 
unit at the left is the FET and bipolar-transistor tester. At the right is seen the crystal and 
bipolar-transistor checker. Various sizes of crystal sockets are installed in order to accommodate 
the popular pin sizes and spacings. 


are crystal sockets with different hole 
sizes and spacings. This feature makes 
the unit more versatile with respect to 
checking crystals in various holder 
styles. Both testers are housed in home- 
made aluminum cases. Fig. 73 shows 
how the testers are laid out. 

Timing and Control Circuits 

There are a number of places in the 
design of amateur equipment where 
timing circuits must be used. These 
include circuits for the control of trans- 
mitters, receivers or transceivers, audio 
side-tone oscillators, antenna switching 
circuits, sweep and control systems for 
SSTV, and even systems for the control 
of repeaters. There are literally dozens 
of ways to design these circuits. Some 
samples are presented in this section. 

Sidetone Oscillators 

One need during the transmission of 
cw is that the operator have a means for 
monitoring his fist. One method is to 
listen to the transmitted signal in the 
station receiver. It allows the operator 
to know the frequency that he is trans- 
mitting on, if he is using a separate 
transmitter and receiver. However, it 
places some constraints upon the receiv- 
er. The muting system must allow the 
receiver gain to be reduced by 80 to 100 
dB, while still delivering a clean tone. 
Alternatively, the operator must be 
willing to accept receiver deficiencies, 
such as clicks generated within the 
receiver, and even a possible frequency 


shift in tire receiver local oscillator, 
caused by strong rf fields. 

A superior approach to cw moni- 
toring is to use a sidetone oscillator. 
Tli is is an audio oscillator that is keyed 
simultaneously with the transmitter. It 
may be activated by rf detection, by the 
dc voltage changes that occur within the 


transmitter, or by a signal from an 
electronic keyer. Many electronic keyer 
circuits have sidetone oscillators built 
into them, along with small speakers. 
The writers prefer systems that inject 
tire sidetone signal directly into the 
audio chain of the receiver. This is more 
compatible with headphone operation. 



Fig. 74 - Circuit of a PUT audio oscillator. 


Test Equipment and Accessories 173 






Shown in Fig. 74A is an audio 
oscillator using a General Electric D-13T 
programmable unijunction transistor 
(PUT). The output frequency is about 1 
kHz. and may be changed by replacing 
the capacitor value shown. The output 
is at low levels, suitable for injection 
into the input of a medium-gain audio 
amplifier. 

'Hie PUT is a device, similar to a 
silicon-controlled rectifier (SCR), that 
can be used for many applications. 
Shown in Fig. 74B is a model for a PUT. 
The three-terminal device may be 
thought of as being composed of a 
combination of an npn and a pnp 
transistor in the form shown. While this 
circuit is used primarily as a model to 
explain the operation of PUTs, one can 
also use this circuit in order to build 
PUT circuits when suitable devices are 
not available. Good choices for the 
devices are the 2N3904 and 2N3906 for 
the npn and pnp, respectively. 

Shown in Fig. 75 is another sidetone 
oscillator consisting of a free-running 
multivibrator which uses two bipolar 
transistors. This circuit will operate with 
virtually any common silicon transistor 
type, and does a good job of generating 
a sidetone. The output is a square wave 
at approximately 1 kHz. The diodes in 
the base are necessary in order to 
prevent damage to the emitter-base 
junctions of the transistors from break- 
down. If the oscillator is run from lower 
voltage supplies (5 volts or less on the 
collector resistors) the diodes may be 
eliminated. 

A variation of this circuit is shown in 
Fig. 76. The circuit uses a voltage that is 
derived from the rf output of a QRP 
transmitter in order to provide part of 
the operating voltage for the circuit. 
This circuit has the characteristic that 
when the transmitter is keyed, the 
output tone occurs. This tells the oper- 
ator that the transmitter is delivering rf. 
Moreover, the pitch of the oscillator is 
proportional to the rf output voltage. 
This means that the transmitter maybe 
tuned in the field without having a 
meter built into the equipment. This 
can be handy when ruggedness and 
minimal weight are design criteria. 

Shown in Figs. 77 and 78 are a pair 



Fig. 75 - Sidetone oscillator using a multi- 
vibrator. 



of oscillators using 741 operational 
amplifiers. The circuit shown in Fig. 77 
is an oscillating variation of a type of 
low -pass audio filter. If the resistor 
values were chosen carefully, it might be 
possible to obtain a fairly clean sine 
wave from tire circuit, although it might 
then be sluggish in starting. The circuit 
of Fig. 78 utilized the 741 as a differen- 
tial comparitor with positive feedback. 
It is generally more predictable than the 
other circuit. 

The oscillator of Fig. 79 uses a 555 
timer IC. These ICs are useful in timing 
applications, and will be discussed later. 



T-R Relay-Control Systems 

In tire construction of cw and ssb 
transmitters (or transceivers), one useful 
accessory is a key (or VOX) controlled 
transmit-receive system. In cw service, 
this means that when tire key is pressed, 
the relay used for transferring the anten- 
na from the receiver to the transmitter 
is activated automatically. Furthermore, 
the transmitter circuit is activated and 
the receiver is muted. In the case of ssb 
operation. Uiese same functions are 
realized with a VOX, or “voice-operated 
switch.” In this case, some audio from 
the speech amplifier is rectified to pro- 
vide a dc voltage that will activate the 
relay -control circuitry. In either the cw 
or die ssb situation, it is desirable that 
the antenna changeover occur quickly, 
and that after the key is released, or the 
voice ceases, the relay stay closed for a 
short period. The length of the hold-in 
time will depend upon the application. 
For contest work, periods around 0.5 
second are suitable. For ragchewing, 
longer periods may be desired. More 
than 1 to 2 seconds is generally avoided. 

Shown in Fig. 80 is a circuit used in 
many stations. This system is com- 
patible with a keying mode that keys a 
positive voltage to ground, the usual 
case with solid-state gear. When the key 
is depressed, the pnp transistor is satu- 
rated immediately. This sends current 
through die base of the npn, which 
activates the relay. The usual practice is 


Fig. 77 — An op-amp sidetone oscillator. 



Fig. 78 - An improved version of the circuit 
in Fig. 77. 



Fig. 79 - An audio oscillator which employs 
an NE555 timer IC. 


174 Chapter 7 













to use a muitipole relay. One set of 
contacts transfers the antenna while the 
remaining contacts apply dc voltages to 
the transmitter circuits. When the pnp 
transistor comes on, part of the output 
current flows into the capacitor through 
the 270-ohnt resistor. This causes the 
capacitor to charge quickly to +12 volts. 
When the key is released, the pnp device 
is immediately cut off. However, the 
timing capacitor is now charged to a 
high potential. The capacitor will dis- 
charge through the potentiometer, 
determining the hold-in time. A diode is 
placed across die relay coil. It protects 
the npn transistor. If die diode was not 
diere, a high positive collector-voltage 
spike would occur at die instant the 
relay turned off. Depending upon die 
inductance and resistance of the relay 
coil, and the stray capacitances, this 
potential could reach several hundred 
volts. The diode clamps this posidve 
voltage spike to the positive power 
supply line. The current that Hows in 
the diode will have the effect of ex- 
tending die hold-in time of the circuit 
slightly. 

The circuit of Fig. 80 lias some 
deficiencies. The main one is that the 
capacitor must be almost completely 
discharged before the relay will drop 
out. The exact time of relay dropout 
will depend on the beta of the npn 
transistor. Beta variations among transis- 
tors of a given type are often large, and 
may be temperature-dependent. 

The deficiencies outlined may be 



Fig. 81 - A differential comparator which 
uses a 741 op amp. 


circumvented by using more precise 
timing circuits. One of the easier ap- 
proaches to such design is through the 
use of a differential comparitor circuit. 
Such a circuit is shown using a 741 
op-amp in Fig. 81. V r is a reference 
voltage that is derived from the power 
supply through a voltage divider. 
Typically V r will be about 0.5 V cc . The 
input voltage of the comparitor is in- 
creased from zero toward the posidve 
supply. As it approaches the reference 
voltage, the output of the 741 will start 
to increase. Since the dc gain of the 741 
is high, the transisdon from the low to 
the high state will occur over a range of 
input voltage of a millivolt, or there- 
abouts. A curve of this response is 
presented in Fig. 81. 

In the example shown, the reference 
voltage is applied to the inverdng input 
of the op amp while the control voltage 
is placed on the noninverting input. If 
die reverse circuit was used with the 
input signal applied to the noninverting 
input, the output would be high for low 
values of input. With 741 op amps the 
output voltage will approach the posi- 
tive supply within a volt or two. Similar- 
ly, in the low state, the output can drop 
down to about 2 volts. The character- 
istic that the output does not come 
closer to ground is sometimes a problem 
that makes additional components 
necessary. Some of the newer op amps 
will allow their outputs to approach the 
supply voltages more closely. An excel- 
lent choice for circuits of this kind 
would be the LM-324, which is a quad 
op amp (four op amps in a single 
package, each with characteristics simi- 
lar to the 741 ). 

A simple T-R control system using a 
741 as a differential comparitor is 
shown schematically in Fig. 82. The 
reference is obtained from a divider, and 
hold-in time is determined by the 
100-ki2 and the 5-pF capacitor. A 5-volt 
Zener diode in the output of the op 
amp assures an output that drops to 
ground potential. If one section of an 
LM-324 was used, the Zener diode 


could be eliminated. A multipole relay 
is used with one set of contacts shifting 
the voltages to the transmitter, as re- 
quired. This ensures that the transmitter 
does not come on until the antenna is 
connected to the transmitter. 

It is important in many cases that 
the antenna relay be in the transmit 
position before rf is applied. If this is 
not done, the relay is required to switch 
when large rf voltages are present. This 
places severe requirements on the relay. 
Furthermore, the transmitter final 
amplifier may be operating for a short 
period with no termination on the 
output. This can lead to instabilities and 
can, in some cases, destroy the trans- 
mitter output transistor. Receiver 
front-end damage is also common. 

Shewn in Fig. 83 is a modified 
system that is designed to circumvent 
these problems. The main relay control 
circuit is identical with drat shown in 
the previous schematic. However, when 
llie key is depressed and the output of 
U1 goes high, a current will flow 
through the 220-kf2 resistor that con- 
nects to U2. The 0.1 -qF capacitor at the 
input to U2 will cause a delay of about 
20 milliseconds before the output of U2 
goes high. The high output at U2 can be 
used to turn on a switch (Ql) that 
grounds the oscillator control. Alterna- 
tively, die output of the switch can be 
used to control a pnp switch (Q2) tiiat 
applies a positive voltage to the oscilla- 
tor in the transmitter. 

At the end of a timing cycle when 
U1 returns to an off condition, the 
oscillator voltage is ternunated quickly. 
Tltis is realized with the diode across the 
220-ki2 resistor. Receiver muting signals 
should be derived from die output of 
Ul. This will ensure that the receiver is 
muted before any rf is generated. 

The 20-ms delay introduced by die 
U2 timing circuit presents a minor 
problem: The first dot of a ew trans- 
mission is a bit shorter dian the signal 
generated by the key. This problem is 



Fig. 82 - A T-R circuit which uses an op-amp 
differential comparator. 


Test Equipment and Accessories 175 






not severe, however, since the length of 
a dot at 20 wpm is about 50 ms. It is 
better to suffer this slight inconvenience 
than it is to burn out a final amplifier, 
or to create a tremendous key click on 
the air when the relay switches while 
“hot’ with rf. This characteristic is 
noticeable with some commercial trans- 
mitters. The 20-ms period was chosen 
because most relays take approximately 
10 ms to pull in. This includes de- 
controlled coaxial relays. The cautious 
experimenter should measure the pull-in 
characteristics of Iris relay with a trig- 
gered oscilloscope, then tailor the time 
constants accordingly. 

While op-amp ICs have been used in 
the previous circuits, they are not the 
only way to handle the relay driver 
problem. Shown in Fig. 84 is a simple 
comparator type of switching scheme 
that offers good timing accuracy. This 
circuit uses two transistors and a Zener 
diode as tire main elements. 

Often it is desired to run an out- 



Fig. 84 - Example ot a relay driver which 
uses iwo transistors and a Zener diode as the 
main elements. 


board power amplifier as an accessory the relay would be used to apply dc fi 

to a low-power transmitter. The best voltage to the outboard amplifier. in 

way to switch the antenna would be to d< 

run appropriate dc control voltages to Circuit Description h . 

the outboard final. However, this would R1 of Fig. 85 serves as an rf voltage bi 

make the accessory less convenient to divider to permit the circuit to be used aj 

use. An alternative approach is that of with transmitters of various power- ar 

using detected rf energy from tire ex- output amounts. Rf energy is routed bi 

ci ter to control a suitable relay in tire through Cl to tire base of broadband th 

outboard amplifier. Shown in Fig. 85 is amplifier Ql. The amplified Ill-band di 

a circuit that was developed for this energy is supplied to a voltage -doubler tc 

purpose. The user should be sure his (CR1 and CR2 ) through a broadband di 

exciter is capable of operation (tempo- toroidal step-down transformer, T1 . The su 

rai'ily) without a load without self- rectified rf voltage at the output of CR1 ct 

destruction. Ideally, a set of contacts on and CR2 is filtered by means of RFC2. 



Fig. 85 - Circuit of an rf-actuaied relay driver. This unit was first described in QST for Aug 
1976. p. 21, inclusive of a pc-board layout. K1 isa 12-V relay with a field coil dc resistance be- 
tween 500 to 1.000 ohms. T1 primary has 25 turns of No. 28enam. wire on an Amidon FT-50- 
43 toroid core. The secondary consists of 5 turns of No. 28 wire wound over the primary. RFC1 
and RFC2 contain 42 turns of No. 28 enam. wire on FT-50-4 3 toroid cores. Fig. 


176 Chapter 7 








Fig. 86 - Block presentation of an NE555 
timer 1C. 

C5, and C6. This prevents unwanted rf 
front reaching U1 and affecting its 
performance. 

C6, R7. and R6 comprise a timing 
network (variable) which governs the 
hold-in time of the relay. Kl. The 
smaller the resistance amount at R6, the 
shorter will be the time delay. 

U1 functions as an inverting ampli- 
fier. When the input dc voltage at pin 2 
increases, the output dc voltage at pin 6 
decreases. The output voltage causes the 
base of relay driver Q2 to be forward 
biased negatively when it drops below 
approximately 1 .4 volts. Diodes CR5 
and CR6, by virtue of their combined 
barrier voltages (0.7 V each), established 
the 1.4-V fixed bias level. Without the 
diodes, Q2 would conduct sufficiently 
to prevent the relay from dropping out 
during no-signai periods. CR4 is used to 
suppress transients caused by the field 
coil of Kl. When no rectified rf reaches 


Ul, Q2 is cut off because of the high 
positive base voltage it receives from 
Ul . and the relay contacts to the 
transmitter are open. 

The NTS 55 Timer 

An IC that is useful for timer appli- 
cations is die NE-555. Several com- 
panies manufacture versions of litis 
chip. The Motorola part number is the 
MC-1555. The principles that are 
applied in (his chip arc similar to those 
described. The chip contains a set-reset 
flip-flop (RSFF). an output buffer that 
will supply or sink up to 200 mA of 
current, two differential comparitors for 
control of die RSFF, as well as some 
other control functions. The typical 
package is an 8-pin mini-DIP. The cir- 
cuit also has a built-in resistive divider 
that provides two reference voltages at 
1/3 and 2/3 of die supply voltage. The 
chip will operate with supplies from 5 
to 18 volts. 

Shown in Fig. 86 is a block presenta- 
tion of the 555 timer chip. The output 
appears at pin 3. Pin 7 can also be used 
as an output. It is an open collector of a 
transistor with a grounded emitter. 
Under most conditions this transistor is 
in an “on" condition when the output. 
Q, at pin 3. is low. The chip is triggered 
into an on condition (Q high ) by pulling 
pin 2 below 1/3 of die supply voltage. 
Pin 4. which is labeled in the literature 
as a reset, serves the function of turning 
on the transistor with output at pin 7. If 
this reset is not to be used, it should be 
tied to the positive supply. The other 




Fig. 88 - An rf-derivnd circuit for operating 
the system shown in Fig. 87. 


reset, which is labeled “threshold,” re- 
sets the RSFF when the terminal 
becomes more positive than 2/3 of the 
supply voltage. Pin 5 is the 2/3 V cc 
reference voltage and should normally 
be bypassed for high frequencies. In 
situations where several 555 timers are 
linked for complex timing functions, all 
of these reference voltages may be tied 
together to ensure accuracy. 

Shown in Fig. 87 is a break-in delay 
circuit using the 555 timer. Under nor- 
mal key -up conditions, the FF will have 
been reset and Q (pin 3) will be low. 
When the key is pressed, the RSFF is set 
into a high condition. This results from 
pin 2 going low. The circuit is inhibited 
from “timing out" by the clamping 
action of CR1. If tlus diode were not 
present, timing would begin as soon as 
die RSFF was set. The timing is pre- 
vented so long as the key is depressed. 
When the key is lifted, the timing 
capacitor, Cl. begins to charge. If the 
key is depressed before the timing has 
finished, the capacitor is discharged 
through the key and CR1 . If the key is 
left open for a period of time. Cl will 
eventually charge to 2/3 V cc . This 
action applied to the threshold terminal 
(pin 6) causes the flip-flop to be reset. 

While the circuit may be used as 
described for simple relay control, a 
simple modification may be made to 
obtain a delayed control for the trans- 
mitter. This is the circuit associated 
with the 741 op-amp. The internal 
reference of the 555 timer is used as die 
reference for the 741 comparitor. The 
delay operation is virtually identical 
with that described for Fig. 83. 

This circuit (Fig. 87) may be modi- 
fied easily to operate from an rf-derived 
signal for use with an outboard ampli- 
fier. This application is shown in Fig. 
88, and the relay is now used in a 
manner identical to that of Fig. 85. 

A somewhat more complex applica- 
tion of the 555 timer is the electronic 
keyer shown in Fig. 89. This circuit is 
straightforward, as keyers go, and the 
performance is good. It has advantages 
over some of the circuits that are 
popular. One is that when a character is 
started (a dot or a dash), no more 
information may be entered into the 



Test Equipment and Accessories 177 







Fig. 89 - An electronic keyer which employs three NE555 timer ICs. 


circuit, irrespective of paddle position, 
until the end of the following space. In 
many circuits it is necessary that the 
user be “off the paddle” before the end 
of the dot or dash. Otherwise, another 
character will be generated. Another 
advantage of this circuit is that the 
capacitors start a timing cycle in a 
completely discharged condition. 
Because of this, it is not necessary to 
discharge the capacitors quickly through 
tiie paddle and additional circuits. This 
phenomenon led to timing errors when 
poor quality components were used in 
an earlier circuit described by one of the 
writers (QST for Nov.. 1971). A final 
advantage of this circuit is that all three 
of the functions (dot, dash, and space) 
are timed with separate circuits. As a 
result, the timing resistors (R1 , R2,and 
R3) may be changed in order to adapt 
to any individual taste. 

The purpose of the foregoing keyer 
description was to demonstrate the 
versatility of the 555 timer. While the 
keyer functions quite nicely and is 
presently in use, there are dozens of 
keyer circuits available that will func- 
tion as well. Undoubtedly, the optimum 
route to follow in such a design would 
be to use CMOS ICs. The power con- 
sumption is very low with such devices, 
and so is the cost. 

The writers have taken a slightly 
different approach to the keyer design 
problem than is perhaps typical. The 
usual approach is very pragmatic, that 
of finding a design of an acceptable level 
of complexity that will provide the best 
performance available. On the other 
hand, keyers offer another profound 
advantage from an educational point of 
view. The function that is to be 
designed is fairly straightforward, and 
yet certainly not trivial. As a result, 
designing keyers is an excellent mecha- 
nism for learning about new circuit 
techniques. Even if the circuits are never 
built, it can be enlightening to go 


through the learning exercise of 
designing them. 

Electronic T-R Switching 

All of the techniques outlined above 
for T-R switching have utilized a relay. 
However, the function can be handled 
completely with electronic switching 
methods. There are two general 
approaches to the electronic T-R switch 
problem. The one that is most common 
is one of attaching the antenna directly 
to the transmitter. Then, the receiver is 
paralleled with the transmitter output 
with suitable circuitry to prevent 
damage to the receiver during transmit 
periods. The other approach is to 
actually switch the transmitted power 
directly. Clearly, this is the more diffi- 
cult of the two. 

For most work on the hf bands the 
simpler method of T-R switching is 
suitable. The advantage of electronic 
T-R switching is that it allows full 
break-in operation on cw,a feature that 
is convenient for the contest operator, 
traffic handler or vhf meteor-scatter 
enthusiast. There are some constraints 
that must be applied to the design of 
the system. First, there should be mini- 
mal degradation in receiver perfor- 
mance. This can originate from two 
considerations. One is that distortion 
products can be generated in the switch, 
which would degrade the dynamic 
range. Furthermore, losses in the switch 
can degrade receiver noise figure. The 
second and major source of problems is 
the transmitter output network. If the 
signal is not extracted from the trans- 
mitter output in the proper manner, 
there may be significant attenuation of 
the signal. Examples of this effect will 
be presented later. Another constraint is 
that the T-R switch not create extra 
harmonic output from the transmitter 
that could cause interference to other 
services. This problem is usually handled 
easily. 


Shown in Fig. 90A is a simple T-R 
switch. The receiver is attached to the 
collector terminal of the matching pi 
network. In this example for 7 MHz, a 
50-pF capacitor is used for coupling 
into the low-impedance port of the 
receiver. The receiver is protected from 
strong rf signals by the back-to-back 



Fig. 90 - Circuit of a simple T-R switch. 


178 Chapter 7 










silicon switching diodes. The 50-pF 
capacitor will become part of the trans- 
mitter tank and reasonable rf currents 
will How here. The diodes must be 
capable of handling this current. 
Because of the reactance in the 50-pF 
capacitor (about 500 ohms at 7 MHz), 
there may be some attenuation of sig- 
nals. This may be avoided with the 
system shown in Fig. 90B. Again, back- 
to-back diodes are used. However, there 
are now two diodes per leg. The 
coupling is into a higher impedance 
point in the receiver input. Because of 
this, the switch presents virtually no loss 
during receive periods. 

There are two major observations 
that should be mentioned about the 
circuits described. First, measurements 
indicate that the back-to-back diodes do 
not cause IMD at the receiver input as 



Fig. 92 - Receiver input protection circuit 
which uses silicon diodes. 


long as the signal is not large enough to 
turn the diodes on. In a 50-ohm system, 
levels at die antenna terminal of up to 
nearly a volt pk-pk could probably be 
tolerated without compromise in dy- 
namic range. The second item of signifi- 
cance is the point of attachment to the 
transmitter. The antenna appears as a 
50-ohm load at the frequency of inter- 
est. presumably. On die Q f side of the 
pi network, a 50-ohm load is also seen. 
However, if the receiver was attached on 
the antenna side of the pi network, the 
receiver would see the 50-ohm resis- 
tance of the antenna in parallel with a 
series-tuned circuit at the operating fre- 
quency. This series-tuned circuit could 
lead to significant attenuation. This 
effect would not be quite as pro- 
nounced in die system of Fig. 90B, for 
some of the reactive effects of the 
series-tuned circuit could probably be 
tuned out by readjustment of the receiv- 
er input tuned circuit. In either case, the 
diodes will create some harmonic cur- 
rents. It is best that the low-pass fil- 
tering action of the pi network be used 
to ensure diat die harmonics never 
reach the antenna, where they may be 
radiated. 

In all of die schemes described, 
silicon switching diodes should be used. 
The relatively low turn-on voltage of 
germanium or hot-carrier diodes would 
cause them to create IMD in the 
received path. 

The examples of Fig. 90 used silicon 
diodes as shunt clamp elements. The 
diodes can also be used in a series 
configuration. An example is shown in 
Fig. 91 where a pair of back-to-back 
series diodes are biased on to a current 


of about 6 mA in each diode. When rf is 
generated by the transmitter, some of 
the output is sampled and rectified. The 
resulting dc is used to saturate 02 which 
has the effect of turning Q1 off. In one 
system of this kind that was investigated 
with a 2-watt QRP transmitter, it was 
found that the receiving insertion loss of 
the switch was about 1 dB, completely 
insignificant at 7 MHz. The attenuation 
of rf from the transmitter was over 40 
dB, which was enough to prevent 
damage to the receiver front end. 
Because the receiver being used had a 
fast, wide-range age system, complete 
break-in operation was possible with no 
clicks or thumps. It’s an unusual 
experience to hear signals between the 
dots in a 30-word-per-minute string. A 
system of T-R switching of this kind is 
used in a superhet transceiver described 
later in the book. That system was not 
set up for QSK operation, however. 

Measurements have not been per- 
formed to evaluate the IMD levels 
created by the series-biased diodes. 
They would probably not be detectable 
unless the receiver had an input inter- 
cept greater than 0 dBm. 

In many cases it may be desirable to 
provide additional protection at the 
input of a receiver from the effects of a 
transmitter. A system shown in Fig. 92 
will do this. It is assumed that some sort 
of a control voltage is available, pro- 
viding + 12 volts when the transmitter is 
on. This signal is used to bias a switch- 
ing diode on to about 6 mA. Since this 
diode is across the hot end of the tuned 
circuit (a high-impedance point), it must 
be reverse biased during receive periods. 
The additional diodes are used to pre- 
vent damage to the FET during switch- 
ing transients. They may not be neces- 
sary. The user should not rely upon the 
Zener diodes that are built into the 
MOSFET front end of the receiver. 
These diodes are typically very small 
and will only handle small currents 
before burnout. They are useful mainly 



Fig. 93 - Circuit for using a pnp transistor 
switch for shaped keying of a stage in a trans- 
mitter. 


Test Equipment and Accessories 179 






Fig. 94 - A pnp keying transistor functioning 
as an emitter follower. 


for the protection of the MOSFET from 
damage during handling. 

While additional measurements are 
required, it appears that methods of the 
kind shown offer great promise for the 
QSK enthusiast. The better results will 
probably come from combinations of 
the methods outlined. The single largest 
factor, other than the obvious effective- 


ness of protecting the receiver, is to 
ascertain the 1MD effects. Spectrum 
analyzer measurements are required. 

Shaped Keying 

A problem with many solid-state cw 
transmitters is key clicks. This is usually 
the result of oversight by the designer. 
So much effort is devoted to the rf 
details of the design that the shaping 
can be forgotten. There are many cir- 
cuits that can be used to assure that the 
cw note is clean and crisp. 

Shown in Fig. 93 is a circuit that is 
used frequently in many of die trans- 
mitters in the book. Here a pnp tran- 
sistor is used as a switch. This circuit has 
several advantages. One is that the key- 
ing is done in the positive supply line, 
but the key is still grounded. This allows 
the builder to carefully ground the rf 
parts of the circuit without regard for 
extra dc control wires. The other virtue 
is that the switch provides an easy 
means of controlling the timing. This is 
performed with the network in the base. 
C'l and R1 determine the rise time of 
the waveform while Cl and R2 control 
the fail time. This circuit is suitable for 
keying stages requiring up to about 50 
mA of current. Greater current amounts 
may be keyed if larger switching transis- 
tors are used. The base resistors must be 
decreased in ohmic value though. 

A Class A amplifier may be keyed 
with the circuit of Fig. 94. Again, a pnp 
transistor is used. However, in this 
application it functions as an emitter 



Fig. 95 - A pnp keying transistor operated as 
an integrator. The timing capacitor, Cl , 
should not be an electrolytic type. 


follower instead of a switch. As such, 
the dc waveform is slightly more pre- 
dictable than with the circuit of Fig. 93. 

Fig. 95 shows a third method for 
applying a transistor to shaped keying. 
In this circuit, the transistor functions 
as an integrator. When the key is closed, 
base current starts to (low. However, 
this causes the voltage on the keyed 
amplifier to begin to rise. The increasing 
voltage is coupled back to the base, 
decreasing the base current. The final 
result is that the collector voltage ramps 
up linearly. A similar action occurs 
during the fall period. While the wave- 
forms are trapezoidal instead of the 
more classic exponentials, they have low 
sideband energy. This results in a clean 
keying characteristic. 


180 Chapter 7 










Chapter 8 

Modulation Methods 


The theory presented in the preceding 
chapters has been general, applying to 
cw and ssb systems. The construction 
projects have been predominantly for 
the cw enthusiast. However, phone is 
the principal mode of operation for 
most amateurs. On the hf bands single 
sideband is predominant. At vhf and 
uhf, there is a split. There is an increas- 
ing number of stations using ssb at vhf 
and uhf. The most common mode is fm. 

In this volume we will treat the 
details of single and double -sideband 
phone transmitters. Frequency modula- 
tion methods are omitted because they 
are covered in detail in many other 
books. 

Our treatment of sideband methods 
will include many problems. The text 
will deal with some introductory infor- 
mation on the design of the component 
parts of a phone transmitter, the design 
of high-power amplifiers, and the vari- 
ous methods that are available for realiz- 
ing these ends. We will attempt to fill in 
some gaps that have appeared in the 
amateur literature. Specifically, the 
design of low- and medium-power Class 
A broadband amplifiers will be covered. 

The Nature of an A-M Phone Signal 

If a cw signal was to be expressed 
mathematically, it would be a simple 
sine wave. That is, the voltage appearing 
at the transmitter output would be V„ = 
A sin to t where w = 27r/is the angular 
frequency. F is the frequency in hertz. 
The term A is the peak amplitude of the 
signal. 

Modulation is a term that implies a 
controlled change. The terms in the 
simple cw signal that may be changed or 
modulated are the amplitude. A, or the 
frequency,/. The amplitude modulation 
(a-m) that is used for standard broad- 
cast, and at one time was the dominant 


phone mode for the amateur, is des- 
cribed by 

V a =Ao(i + fcsin 2rr/ m /)sin 2nf c r 
= Ao sin 2itf c + A„k (sin2 nf m t) 

(sin 2 rr/ c r)- fFo 11 


There are two frequencies repre- 
sented. F c is the carrier frequency. This 
is evident from the expanded form of 
the equation where we see that there is 
a steady output at the carrier frequency. 
A 0 is the peak carrier amplitude, and 
the term k is called the modulation 
index. 

The other term in the expanded 
form of Eq. 1 is a product of two sine 
waves. The two sine terms have two 
frequencies, f c and f m . The second 
frequency, / m , is the modulation fre- 
quency. If we refer back to the dis- 
cussion of mixers and product detectors 
in the receiver chapters, we will recall 



Fig. 1 - Time-domain oscilloscope presenta- 
tion of an a-m phone signal. 100 percent 
modulation is present. The modulating fre- 
quency is 1 kHz. 


that a device which has an output that is 
a product of two input voltages (a 
multiplier circuit) has as its output a 
pair of signals at the sum and difference 
frequencies. Hence, the total output of 
the a-m transmitter will contain three 
frequencies. One is the carrier, f c . Tire 
other two are called the sidebands, and 
are at frequencies f c + f m and f c - f,„ . 
Shown in Fig. 1 are oscilloscope presen- 
tations of an a-m signal. In Fig. 2 are 
spectrum-analyzer presentations of the 
same signal. The carrier and the side- 
bands are clearly evident. In the case 
shown, the carrier frequency is 432 
MHz. The modulation frequency is 1 
kHz. If the value of k is multiplied by 
100, the result is the percentage of 
modulation. The signal shown in the 
photographs is modulated 100 percent. 

It is interesting to note the powers 
that are associated with the various 
frequencies in a 100-percent modulated 
carrier when a single sine wave is used as 
the modulating signal. From Eq. 1 we 
see that the carrier power is a constant. 
The voltage of the carrier is A o_ volts, 
peak. The rms value is A a -r \/T7 Since 
the power is delivered to a resistive load. 
R, the power is just V 2 -r R , or.4„ 2 -r 
2 R. 

If Eq. 1 is expanded, using trigono- 
metric identities, we see that die ampli- 
tude of each of the side bands is A,, + 2 . 
I lence. the average power in each of these 
sidebands is 0.25 of that in the carrier. 
A spectrum-analyzer display of an a-m 
signal which is modulated 100 percent 
by a single audio frequency will show 
accordingly that the average power in 
each of the sidebands is 6dB below the 
carrier power. 

If we go back to Eq. 1 , we see that 
die normal cw signal with an amplitude 
of A n is replaced by one with a variable 
amplitude. At some parts of die audio 


Modulation Methods 181 




Fig. 2 - Frequency-domain presentation ot 
the a-m signal of Fig. 1. The spectrum 
analyzer was set for a vertical display of 
2 dB per division. Note that the sidebands 
are 6 dB below the level of the carrier. 

The modulating frequency is 1 kHz and 
the carrier is at 432 MHz, 


cycle, the instantaneous amplitude is 
zero. At other parts, where the audio 
oscillation is at its peak value, the 
amplitude of the sine wave is twice as 
high as that of the carrier alone. This 
factor of 2 in voltage results in a factor 
of 4 in power. This power is called the 
peak-envelope power (PEP) and is 6 dB 
greater than the carrier power. 

In the foregoing discussion, it has 
been assumed that the modulation 
signal is a single-frequency sine wave. 
This makes the mathematics easy. In the 
real situation, the audio signal would be 
the voice of the operator. This is com- 
posed of a number of sine waves added 
together to form a composite voice 
waveform. The transmitter will behave 
essentially as if each of the component 
sine-wave modulating signals were 
present alone. Then, the net output 
would be the addition of each of the 
individual components. 

The Double-Sideband Signal 

If the a-m signal is studied with a 
spectrum analyzer or mathematically, 
we find that the carrier at f c has a 
constant level. As the audio signal is 
applied to the transmitter, the levels of 
the two sidebands vaiy, but the level of 
the carrier remains constant and un- 
altered. Hence, it contains no informa- 
tion. It is necessary if the signal is to be 
detected in a receiver using a simple 
rectifier detector, but it serves no oilier 
purpose. On the other hand, when we 
examined the average power in the 
carrier and in the two sidebands of an 
a-m signal, we found that most of the 
power was in the carrier. It would be 
much more efficient if we could concen- 
trate all of the transmitted power in the 
sidebands where voice information is 
contained, and dispense with the carrier. 
This can be done: The result is a double- 
sideband signal. 

If the spectrum-analyz.er photo- 
graphs of an a-m signal are studied, a 


double-sideband signal can be en- 
visioned: It is the same presentation 
without the carrier present. This is done 
with a balanced modulator in practice. 
This circuit, which will be discussed in 
more detail, is essentially a balanced 
mixer. One of the inputs is at audio 
while the other accepts the carrier fre- 
quency. 'Hie output is balanced so that 
the carrier does not appear in the 
ou tput. 

The output of a double-sideband 
transmitter differs from the a-m one, in 
that there is virtually no rf output 
present until an audio tone (or voice 
waveform) is applied to the modulator. 
Then, rf output will occur. The average 
power in each of the individual side 
bands is always equal. 

The Single-Sideband Signal 

If the double-sideband signal is 
investigated, we see that each of the two 
sidebands is of equal amplitude and 
each contains the same information. 
Because of this, an improvement in 
efficiency can be obtained by removing 
one of die sidebands. With only one 
sideband being transmitted, all of the 
available power can be concentrated in 
tlie remaining one. Shown in Fig. 3 is a 
collection of spectrum sketches. A 
combination of three audio tones (Fig. 
3A) is impressed simultaneously on a 
carrier. Fig. 3B shows the result with an 
a-m transmitter. Fig. 3 C shows the 
result when a double-sideband trans- 
mitter is used. The spectrum that would 
result from these tones being trans- 
mitted on a single-sideband (ssb) trans- 
mitter is presented in Fig. 3D. 

It is interesting to consider the 
power related to a single side-band 
transmission. Consider the usual case 
that is used for the testing of an ssb 
transmitter where two equal audio tones 
are applied to the audio input of the 
transmitter. The resulting output is 
shown in Fig. 4. Fach of the tones has a 
given power associated with it. The 
average total power is merely the sum of 
these two, or twice the value of the 
individual signals. The peak-envelope 
power, however, is four times the value 
of each of the individual tones. The 
reason for this difference is because the 
two audio tones are not related to each 
other (they are incoherent). Because of 
this, there will be instants during the 
transmission when the individual equal 
voltages from each tone are both at 
their peak values simultaneously. The 
net voltage at the output at the instant 
is twice die value of one of die tones, 
and die resulting peak-envelope power 
(PFP) is 6 dB above the power in each 
of the tw o tones. 

In a practical case it is much more 
difficult to relate die average power of 
an ssb signal to the PEP value. This will 
depend upon the characteristics of die 



AUDIO SPECTRUM 
IA) 



fc .CARRIER FREQUENCY 
A-M SIGNAL 
(B) 


FREQUENCY 

V ) 

V 
USB 

DSB SIGNAL 
(C) 



USB 

SSB SIGNAL 
(D) 


Fig. 3 — Representative spectrum displays for 
various modulation forms. The audio spec- 
trum of three tones is shown at A. B shows 
the result when this audio signal is applied to 
an a-m phone transmitter. C shows the out- 
put spectrum when the three audio tones are 
applied to a suppressed-carrier double- 
sideband transmitter. At D. the output spec- 
trum of an ssb transmitter is presented. Note 
that the ssb signal is exactly the same as the 
audio input except that it is translated in 
frequency. 

voice that is being transmitted and upon 
die nature of the transmitter. Some 
transmitters use speech clipping or 
processing in order to limit die peak 
value of the waveforms while increasing 
die average power. In most cases where 
such methods are not employed, die 


i — *->/ 

LSB ,e 


182 Chapter 8 








(X 

UJ 

S 

£ 


0 II 12 

AUOIO FREQUENCY 

AUDIO INPUT 
(A) 



(B) 


Fig, 4 — Spectrum obtained during two-tone 
testing of an ssb transmitter. At A the audio 
input is shown, consisting of two equal audio 
tones of identical amplitude. At B is shown 
the resulting ssb output including third-order 
mtermodulation distortion products. Note 
the frequency spacings of the IMD products. 

PEP value will be much greater than 
the average power of an ssb signal. 

Single -Sideband Generation 

There are two general methods that 
are commonly used for the generation 
of ssb. One is the filter method and the 
other is the phasing method. A block 
diagram of a filter type of ssb generator 
is shown in Fig. 5. This technique is 
virtually identical to that used in a 
superheterodyne receiver, except that 
the signal direction is different through 
the transmitter than it is in the receiver. 
An audio signal is obtained from a 
microphone and is amplified in a speech 
amplifier. It is then applied to the input 
of a balanced modulator. The output of 
the balanced modulator is a double- 
sideband signal. The carrier for the 
balanced modulator is most often 
obtained from a crystal-controlled 
oscillator. 

The dsb signal from the balanced 
modulator is applied to a crystal or 


mechanical filter. This filter is designed 
such that one of the sidebands from the 
modulator is within the passband while 
the other is not. The result is an ssb 
signal. For high-quality ssb signals to be 
generated it is not necessary that the 
filter response have a symmetrical 
shape. It is only necessary that the 
suppression be quite good for the un- 
wanted side band. If a symmetrical filter 
is used, as is usually the case, the crystal 
in the carrier generator (used to drive 
the balanced modulator) may be 
switched, allowing the operator to 
change the sideband that is being 
transmitted. If the sideband from the 
filter is higher in frequency than the 
carrier, it is called the upper sideband 
(usb). The lower sideband (lsb) is simi- 
larly defined. 

Since fixed-frequency filters are 
usually employed for the generation of 
ssb, it is necessary that the intermediate 
frequency ssb output be heterodyned to 
the frequency of interest. This is done 
with a mixer and 1.0 system. Again, we 
emphasize that the filter method is an 
exact analogy to the superheterodyne 
receiver. Either single or multiple con- 
version may be employed. 

The second method used for the 
generation of ssb is called the phasing 
method. This is shown in Fig. 6. The 
basis of such a ssb generator is a pair of 
balanced modulators. Each is driven 
with identical carrier frequencies and 
audio signals of identical amplitude. 
However, the phase of the signals is 
different. The carrier signal driving one 
modulator is 90 degrees out of phase 
with that driving the other. Similarly, 
the audio driving one balanced modula- 
tor is 90 degrees out of phase with that 
driving the other. The outputs of the 
two balanced modulators are added 
together with the result that only one of 
the sidebands remains. It is not immedi- 
ately obvious that such a collection of 
circuits will lead to a single sideband. 
However, the mathematics used to show 
that this does occur are straightforward 
and are presented in the appendix. 

In the early days of amateur ssb the 
phasing method of generation was popu- 
lar. The reason for litis was that the 



Fig. 6 - Block diagram showing the phasing 
method of ssb generation. Two balanced mod- 
ulators are used, each being driven with rf and 
audio signals of identical amplitudes. The rf 
and audio signals to the two modulators are 
in phase quadrature. A mathematical analysis 
is presented in the appendix. 


technology for filter construction was 
not as advanced as it is today. Further- 
more, the phasing method may be 
applied directly at the band of interest. 
A superheterodyne approach to design 
is not mandatory, although it may 
certainly be used. 

Today, the situation is reversed. The 
filter method is predominant for side- 
band generation. This is largely a result 
of the nature of the filters that are 
available, along with the transceive con- 
cept where the same filter may be used 
for sideband generation and to obtain 
receiver selectivity. The other reason is 
that the filter method does not exhibit 
the fundamental disadvantages that are 
typical of the phasing method. This 
requires some elaboration. 

If a single audio tone was to be 
transmitted at a single frequency with 
the phasing method of ssb. the design 
would be straightforward. Building net- 
works that provide 90 degrees of phase 
shift at a single frequency is generally 
easy. This is not what is needed for 
sideband generation, though. The rf 
phase-shift network must operate accur- 
ately over a small range, equal in the 
worst case to the width of a phone 
segment of an amateur band. This is not 
difficult to realize in practice. What is 
difficult is the construction of the audio 
phase-shift network. The voice spectrum 
is generally considered to be from 300 
to 3,000 Hz. This is a ratio of 10 in 
frequency. It is difficult to build phase- 
shift networks that will maintain a 
90-degree phase difference with con- 
stant output amplitude over this large 
range. It can be shown that as little as 
r>ne degree of error in the audio phasing 
will lead to an ultimate suppression of 
the undesired sideband of only 41 dB. 

Technology is changing and modern 
methods may inspire a renewed interest 
in die phasing types of ssb generators. 



frequency is adjusted so that it coincides with a point that is 20 dB down on one of the 
sides of the response of the bandpass filter. After an ssb signal is obtained at an intermediate 
frequency it must be hetrodyned to the desired output frequency. 


Modulation Methods 183 









♦3V +3V 




4 


l_2£ 1 

Cl 01 

m r\ c 


■ t Q 

-o U, * 5 

r- 


r 

U2A 

01 01 

] 


A7 


L 

X 





[ 

01 01 

— — n.t 

E. 




2F 


U2B 

Cl 01 











Fig. 7 — A circuit using digital ICs for genera- 
tion of quadrature rf signals as the drive for 
the balanced modulators in a phasing ssb gen- 
erator. Nominally, the ICs would be TTL 
D-type flip flops such as ihe SN7474. For 
higher frequency operation, suitable MECL 
equivalents could be used. 


Radio-frequency phase shift can be 
achieved easily using digital methods 
which are inherently broadband. Speci- 
fically, if two quadrature (90-degree 
phase difference) outputs are desired at 
a frequency f c < one starts with an 
oscillator at 4 f c . This signal is then 
applied to a digital divider using a 
flip-flop with complimentary outputs. 
The result will be two output signals at 
a frequency of 2 f c which are 180 
degrees out of phase with each other. 
Each of these signals is applied to 
flip-flop dividers. The resulting outputs 
will be at the desired f c and will be in 
quadrature. A slightly more elaborate 
interconnection of digital ICs will be 
required than that described, in order to 
ensure that the proper sideband will 
result every time power is applied. This 
is shown in Fig. 7. 

The other phase-shift problem which 
is being changed by modern technology 
is the one occurring at audio fre- 
quencies. The classic circuits that were 
used for audio phase shifting contained 
resistors and capacitors in a complex 
network. The newer approach embodies 
an active phase-shift network. A circuit 
is shown in Fig. 8, where resistors and 
capacitors have been combined with an 
operational amplifier to obtain a phase- 
shifted output. High-performance ver- 
sions of this method will use a multi- 
plicity of these active networks (cas- 
caded) in order to obtain accurate phase 
shifts over a wide range of audio fre- 
quencies. No component values are 
given in Fig. 8 since they will depend 
upon the accuracy desired. The reader 
who is interested in studying this design 
technique should investigate the 1970 
paper by F. R. Shirley (see the bibliog- 
raphy). Using quad op amps like the 
LM324, builders should be able to 
make the phase-shift networks compact 
and low in power comsumption. If a 
phasing ssb exciter is to be built, it is 
important that the audio signals reach- 
ing the phase-shift networks be carefully 


confined to the spectrum of interest. 
Because of this requirement, the speech 
amplifier should include extensive filter- 
ing. The RC active filters discussed in 
connection with receiver design may be 
used to realize this end. 

The writers have not used any of the 
technology outlined for the construc- 
tion of phasing exciters. Our work has 
been confined to the filter approach 
and to double-sideband transmission 
methods. 

While the filter and the phasing 
methods of sideband generation are 
familiar to many amateurs, there are 
others that may be used. One is known 
as "the third method,” or Weaver 
method, named after its originator. A 
reference is given in the bibliography for 
this technique. Also, it may be shown 
mathematically that a carrier which is 
amplitude modulated properly and fre- 
quency modulated simultaneously will 
yield a single-sideband output. 

Single- and Double-Sideband Receivers 

Receivers for single side-band are 
usually "superhets.” That is, they 
employ the filter method for reception. 
However, the phasing method or the 
Weaver approach may be applied to ssb 
reception. There has been some recent 
work with both of these, which are 
essentially extensions of direct- 
conversion receivers. While each method 
works, both have limitations. The main 
problem with the phasing method for 
receivers is the limited sideband sup- 
pression available. A sideband suppres- 
sion of 40 dB is acceptable in the ssb 
transmitter. However, this level would 
be intolerable in any but the simplest of 
receivers. Furthermore, the complexity 
of die filter method is so much less than 
a phasing approach to receiver design 
that we do not recommend the phasing 
technique. One exception would be 
those cases where extremes of sideband 
suppression are not needed. For 
example, one might use the phasing 
method in a receiver as a technique for 
filtering the i-f amplifier for noise. This 
would replace the matched noise filter 
that might be used between the i-f 
amplifier and the product detector in an 
advanced receiver. The main selectivity 
of the receiver is still provided by a 
high-performance crystal filter at the 
input to the i-f amplifier. References are 
given in the bibliography. 

Direct-conversion receivers may be 
used for the reception of single side- 
band. The only problem encountered is 
the audio image. This image frequency 
may contain another station that would 
cause interference to the desired one. 

Double-sideband reception is straight- 
forward with the typical superhet 
receiver. This is because the filter in 
the receiver removes one sideband, 
converting the signal arriving at the 


product detector to an ssb signal. For 
this reason, dsb transmitters are compa- 
tible with stations operating ssb. Indeed, 
the operator may not realize that the 
other sideband is present. However, the 
presence of the other sideband could 
cause interference to other stations. For 
this reason, we do not recommend dsb 
transmitters for use in crowded bands. 

A problem is incurred when one 
attempts to receive a dsb signal with a 
direct-conversion receiver. Since the 
operation of a “d-c” receiver is essen- 
tially that of heterodyning the energy in 
the rf spectrum directly down to audio, 
proper dsb operation detection occurs 
only when the receiver BFO is exactly 
at the same frequency, and has the same 
phase as the suppressed carrier of the 
dsb transmitter. This can be realized 
through advanced detection methods, 
but is generally not recommended. 
Again, the reason is the circuit compli- 
cation and the extra spectrum occupied 
by the dsb signal. 

Balanced Modulators 

All of the techniques used for the 
generation of dsb and ssb use a balanced 
modulator. There are numerous methods 
for realizing such a circuit. Some of 
them will be presented in this section. 

A balanced modulator is nothing but 
a balanced mixer. These circuits have 
been discussed in detail in connection 
with receiver applications. The differ- 
ence between the ordinary receiver bal- 
anced mixer and a balanced modulator 
is in the frequencies presented to the 
input of the circuits. The receiver mixer 
has two radio frequencies at its input, 
with an intermediate-frequency output. 
The balanced modulator has one radio 



Fig. 8 — Circuit showing an RC active audio 
phase-shifting circuit. This is an "all pass" 
network, with the output-voltage amplitude 
equal to that at the input. However, the out- 
put will be phase shifted. In practice, a pair 
of chains of such circuits will be employed. 
Each chain will contain from two to four 
cascaded circuits of the type shown. The 
inputs of the two chains are driven in paral- 
lel. The two resulting outputs are applied to 
the balanced modulators. For calculation of 
the values of R1, R2. R3, R4, Cl and C2, the 
reader is referred to the engineering literature 
(F.R. Shirley, Electronic Design, Sept. 1, 
1970). The op amps may be a 741 . one half 
of a 747 or 5558. or one quarter of anLM324 


184 Chapter 8 






Pig. 9 — A balanced modulator using the MC1496G. The bO-kS2 control is adjusted tor opti- 
mum carrier balance. 


frequency and an audio input. The 
outputs are the sum and difference 
frequencies, or the two sidebands. The 
balance in the circuit ensures that a 
minimum amount of carrier energy 
feeds through to the output. Represen- 
tative values of carrier balance or sup- 
pression are from 30 to 70 dB. For an 
ssb transmitter using the filter method, 
carrier suppressions of 50 dB or greater 
are sufficient. This is because the filter 
will often add another 20 dB of carrier 
suppression. 

The operating power level of a bal- 
anced modulator is somewhat critical. 
As outlined, a voice waveform can be 
analyzed as a composite number of sine 
waves. If the balanced modulator is 
operated at levels that are too high, 
intermodulation distortion will occur 
between these components to make the 
voice sound distorted. If the balanced 
modulator is used in a filter type of ssb 
exciter, all of the resulting distortion 
products reaching the antenna will be 
within the voice spectrum. This is be- 
cause the filter will remove the unde- 
sired ones. However, in a phasing ssb rig 
or in a dsb transmitter, some of the 
distortion products could lie outside of 
the desired voice sideband. 

Shown in Fig. 9 is the circuit for a 
balanced modulator using the MC1496G 
1C. A potentiometer is used for adjust- 
ment of the balance. With careful 
setting of this control, a carrier suppres- 
sion of 60 dB is achieved easily up 
through 10 MHz. An easy way to adjust 
this control is to listen to the mixer 
output in a receiver, then set the control 
for minimum output (no audio applied). 
Using this circuit, tire recommended 
output level is around -10 dBm. If it is 
desired to operate at high output levels, 
the current standing in the 1C should be 


increased. This is done by changing the 
10-kfJ resistor leading to pin 5 to a 
3,300-ohm unit. In this case the maxi- 
mum output should be around 0 dBm 
or less. The recommended output levels 
are for each tone during a two-tone test, 
where a single audio signal is placed at 
the input. With the levels suggested, 
IMD products should be below the 
output by 20 dB or more. This is 
probably adequate for ssb exciters using 
tire filter method. Phasing ssb exciters 
and simple dsb transmitters using this 
circuit should be run at lower output 
levels. 

A similar circuit is shown in Fig. 1 0, 
where an SN76514 is used. Although 
not shown in the literature for this 
device, the carrier suppression may be 
improved with the addition of a control, 
as shown. The recommended output 
levels for this circuit are about the same 
as with the MC1496G. 

A number of balanced-modulator 
circuits are available to the builder who 
uses diodes. Shown in Fig. 1 1 A is one of 
the simplest of these. It has but two 
diodes. In this circuit the balance will 
vary with frequency and is dependent 
primarily upon the match in the diodes 
and the symmetry of the transformer. 
The recommended carrier-oscillator in- 
jection power for all of the diode circuits 
shown is +13 dBm. At this injection 
level, the circuit may be operated at 
output powers up to 0 dBm per tone in 
a two-tone output (which results from a 
single audio input tone). 

Some variations of this circuit are 
also shown in Fig. 11. One uses a 
variable resistor in series with the diode 
pair, with the output being obtained 
from the arm of the control. This circuit 
is recommended for use on the lower hf 
bands and is capable of providing a 


Fig. 10 — The SN76514 mixer 1C used as a 
balanced modulator. The SN76514 has been 
reidentified as TL-442-CN by Texas 
Instruments. It may be procured under either 
part number. 



Fig. 1 1 Balanced modulators using two 
diodes. These circuits are ideal for the 
construction of simple dsb transmitters (see 
text for a discussion of components). 


Modulation Methods 185 







carrier suppression of up to 50 dB, if 
carefully adjusted. The other method 
for balance adjustment (Fig. I 1C) uses a 
pair of variable capacitors. This tech- 
nique is best for vhf applications. We 
have measured over 50 dB of carrier 
suppression at 144 MHz with this cir- 
cuit. The two methods could be com- 
bined for an improvement in suppres- 
sion at the lower frequencies. 

The choice of diode type will 
depend upon the frequency of opera- 
tion. For vhf applications, a hot-carrier 
diode is suggested. However, for the hf 
bands suitable results could be realized 
with 1N914 or similar types of silicon 
switching diodes. In most of the circuits 
presented the audio signal is introduced 
at the center tap of the transformer. It 
is possible to apply the audio directly at 
the connection between the diodes. This 
is realized with an rf choke to isolate 
the rf output from the audio system 
(see Fig. 1 1 D). This may lead to slightly 
improved balance at uhf and could be 
the recommended circuit for building a 
432-MI Iz dsb exciter. 

Shown in Fig. 12 are two other 
diode balanced modulators. Those cir- 
cuits with four diodes are doubly bal- 
anced, although it is not a necessity in 
this application. With any of the diode 
balanced modulators the output should 
be terminated in 50 ohms on a broad- 
band basis. It may be useful to employ a 
low-pass filter at the output of the 
modulator to reduce the harmonic con- 
tent, especially when dsb transmitters 
are being built. Prepackaged diode-ring 
mixers are not recommended, since 
there is no way to adjust carrier suppres- 
sion. 

If careful design work is intended, 
the data presented in connection with 
mixers for receivers should be con- 
sulted. Specifically, the intercept at the 
output port should be studied in order 
to determine the level for proper opera- 
tion of the mixer. The higher output 
levels have the advantage that less gain is 
needed in the following stages. This can 
be a major advantage in a double-side- 
band transmitter. On the other hand, in 
a filter type of ssb exciter, gain is 
achieved in an i-f amplifier. This means 
that the balanced modulator can be 
operated at a low level to make distor- 
tion effects inconsequential. The output 
should not be reduced too far though. 
This could raise the broadband noise 
output of the transmitter. 

In all of the balanced modulators 
shown, the audio port is dc coupled. As 
a result, a cw output can be produced 
by injecting a dc voltage to unbalance 
the modulator. If the carrier suppression 
is good, the transmitter may be keyed 
by shaping the dc that is applied. In 
most situations it will be desirable to 
key an additional stage in the trans- 
mitter. Examples are presented later. 


An additional advantage of the 
dc-coupled nature of the audio-input 
port is that a-m phone operation may be 
realized. A slight amount of carrier is 
inserted by injecting a dc component of 
current until the proper levels are 
obtained at the output. The output 
should be monitored on an oscilloscope 
until 100-percent modulation is 
obtained. Methods are outlined in The 
ARRI. Radio Amateur's Handbook. 

I-F Amplifier and Transmit-Mixer Design 

Wi tit a few exceptions, the design of 
the i-f system for a filter type of ssb 
exciter parallels the same section of a 
superhet receiver. The differences are in 
the output level of the mixer desired, 
and the level that may be applied to the 
crystal filter. 

As mentioned in the previous discus- 
sion of balanced modulators, in a filter 
ssb system the output of the modulator 
may be kept to a low level. This 
minimizes distortion in that circuit. The 
additional gain is then obtained in tire 
i-f system. There are upper limits on the 
signal level that should be reached 
within tire i-f. First, it is sometimes 
dangerous to crystal filters if tire power 
level impressed at their input is exces- 
sive. This will, to some extent, depend 
upon the nature of the filter. With most 
units designed for ssb bandwidtlrs, levels 
as high as 10 to 100 mW will not cause 
damage. The real problem conres with 
narrow-bandwidth crystal filters, as 
might be used for some cw applications. 
This only becomes of significance in the 


present discussion of ssb methods if the 
builder is considering a multimode 
transceiver. 

The main constraint on power levels 
within the i-f section of an ssb exciter is 
in the level used to drive the mixer. In 
our discussion of receiver mixers, we 
found that there was a wide variety of 
performance available. Specifically, vari- 
ous mixers were capable of different 
output intercept values. The transmit 
mixer that follows the transmitter i-f 
amplifier should be operated such that 
the distortion is minimized. Generally, 
this implies that the IMD from the 
output of the mixer should be at least 
40 dB below the desired outputs in a 
two-tone test. This means that the 
output of each tone should be at least 
20 dB below the output intercept of the 
mixer. On the basis of measurements 
that we have done, this suggests that the 
mixer output should be around 5 dBm 
for diode-ring mixers, and should be 
10 dBm or less for an MC1496 mixer. 
This assumes that the MC1496 is biased 
for optimum signal-handling capability. 

Shown in Fig. 13 is an i-f amplifier 
using bipolar transistors. It is followed 
by a MCI 496 mixer. This circuit is 
designed around a KVG crystal filter 
which has an input and output termina- 
tion requirement of 500 ohms. The first 
stage in the amplifier has a variable gain. 
This is realized with a variable resistor in 
the emitter circuit of the stage. Note 
that the current in the transistor is kept 
constant to maintain a high signal- 
handling capability. The second stage 



Fig. 12 - Balanced modulators using diode rings. The 250-ohm control in B is adjusted for 
optimum carrier balance (see text). Also see the previous discussion of product detectors 
and mixers using diodes (chapters 5 and 6). 


186 Chapter 8 




also employs emitter degeneration. The 
main need for this is to maintain a high 
input impedance to the amplifier. 
Because of the light loading that the 
amplifier presents, the termination on 
the crystal filter is determined by the 
external 510-ohm resistor. The output 
of the amplifier is applied directly to 
the mixer input, while the LO port of 
the mixer is driven by a suitable VFO. 

Field-effeot transistors may also be 
used in the i-f amplifier. A transmitter 
presented later in die chapter usesdual- 
gate MOSFETs in the i-f section. 

While there are a large number of 
mixer devices that may be used in 
high-level transmit applications, it is 
highly recommended that a doubly bal- 
anced design be chosen. The section on 
the discussion of transmit mixers given 
in an earlier chapter should be consul- 
ted. Generally, we would suggest that a 
MC1496 be used for single-band designs 
up through 30 MMz. This IC is easy to 
apply and offers suitable, if not spec- 
tacular signal-handling capability. For 
use into the vhf spectrum, a diode type 
of doubly balanced mixer is recom- 
mended. This would also be ideal for a 
multiband hf design because of the 
broadband capability of the circuit. 
However, it is important that the proper 
levels be maintained throughout the 
system. The measurement of low levels 
of rf power was discussed in chapter 7. 
It is recommend that the designer use a 
low-level detector (such as the square- 
law detector described earlier) in con- 
junction with a step attenuator in these 
projects. 

When using a diode-ring mixer, all of 
the precautions about termination 
detailed in the receiver chapter should 
be followed. In a single-band design it is 


often possible to use diplexer circuits, as 
were presented. However, a much 
simpler approach is to use a 6-dB 
attenuator with a characteristic impe- 
dance of 50 ohms at the output. This 
was not desirable for the receiver 
because of noise-figure degradation. 
However, noise figure is of less signifi- 
cance in transmitter applications. The 
6-dB pad should ensure that all mixer 
products are terminated properly. 
Correct LO injection should also be 
employed for the mixer. For diode 
rings, this is from +10 to +13 dBm to a 
50-ohm load. 

The mixer should be followed with a 
bandpass filter. The complexity of this 
filter will depend upon the exact 
frequencies involved. The main spurious 
response to guard against is the image. 
For example, if a 9 -MHz i-f were used in 
a single-conversion transmitter for the 
50-MHz band, the required LO fre- 
quency would be 41 MHz. The image 
frequency would be 32 MHz. A double- 
tuned circuit would provide more than 
sufficient rejection of this component. 


Three-pole filters might be more desir- 
able for most of the hf bands. The 
filters listed in the appendix are suitable 
for this application. 

In some cases, a low -pass filter might 
suffice. For example, if a transmitter 
was built for the 75-meter band, with an 
i-f of 9 MHz, the LO would probably be 
at 5 MHz. If the balance of the mixer is 
reasonable, the 5-MHz output compon- 
ent will be attenuated considerably 
prior to filtering. The main spur would 
be the image at 14 MHz. A low-pass 
filter with a 4-MHz cutoff frequency 
would provide more Ilian sufficient sup- 
pression. The better circuit would 
include a trap or two with frequencies 
of high attenuation near 5 MHz. This 
would provide additional attenuation of 
the LO than might result from less than 
optimal mixer balance. 

Dual-conversion systems should be 
avoided. The high signal levels that are 
often present can lead to distortion 
effects. These are complicated with 
extra conversions of the signal. A better 
approach would be to premix a low- 



Fig. 14 - Circuit of an rf amplifier that might follow the mixer in Fig. 13. Band switching is 
simplified by multiplexing the dc voltage for the amplifier on the output signal line. A typical 

gain for this circuit would be 20 dB, with an output intercept of +20 dBm. 

Modulation Methods 187 









Fig. 16 — Examples of amplifiers with shunt feedback and emitter degeneration. T1 is a broad- 
band transformer on a ferrite core with an N:-1 turns ratio. 


frequency local oscillator with a crystal- 
controlled source. This output would 
serve as a suitable LO injection for the 
single transmit mixer. This method can 
be followed on all amateur bands up 
through 432 MI Iz if a 9-MHz i-f is used 
and advanced filter design methods are 
employed. These filters are difficult to 
fabricate in the vhf and uhf region, but 
are certainly possible. 

Broadband Class A Amplifiers 

In the previous section limits were 
placed on the maximum output power 
that should be obtained from a transmit 
mixer. These contraints resulted from 
the need to keep intermodulation dis- 
tortion to a minimum. The level for an 
MCI 496 was around —10 dBm. By the 
time we add in the loss of the bandpass 
filter, levels as low as -15 dBm might 
be available. While diode -ring mixers can 
provide output powers which are some- 
what higher, much of this extra power is 
absorbed in the 6-dB attenuator recom- 
mended for proper mixer termination. 

On the other hand, most of the 
higher level Class AB amplifiers that are 
used for ssb service require a drive 
power of 1 to 5 watts. To reach this 
level. 45 or 50 dB of gain are required 
following the mixer. While this is not 
difficult to obtain, the problems be- 
come more severe when distortion 
requirements are considered. 

One solution is to use narrow-band 
circuitry. This would not be out of line 
for part of the output chain of a 
band-switched exciter. An amplifier 
could be imbedded within each of the 
filters in order to provide gain. Shown 
in Fig. 14 is a filter of this kind, with a 
dual-gate MOSFET amplifier included. 



Fig. 15 - Example of shunt feedback in a 
broadband Class A medium-power ampli- 
fier. This circuit will provide a gain of 20 dB, 
outpower power of +23 dBm PEP and an out- 
put intercept of +37 dBm. 

T = 5 blfilar turns, 0.2-In, ferrite core, u, = 
850. 

L = 16 turns on T-37-6 (0.77 pH). 


Note that no additional band-switch 
wafer would be required for this circuit, 
since the power supply is multiplexed 
onto the output of the circuit. The 
input triple-tuned circuit is one from 
the catalog of filters in the appendix 
and the output is a single, broadly tuned 
circuit. This stage should provide up to 
20 dB of gain on all of the hf bands 
with an output intercept power of 
approximately +20 dBm. To ensure that 
the intermodulation distortion contri- 
bution from this stage is kept to a level 
of 40 dB or better, the output power 
should not exceed 0 dBm. 

One could continue with a narrow- 
band amplifier design. This would be 
ideal in the case of a single-band trans- 
mitter. However, if the transmitter were 
to be used on several bands, a better 
solution would be to use broadband 
circuitry. The spectrum of signals arriv- 
ing at the input to such an amplifier is 
now well defined. The distortion in a 
broadband amplifier may cause inter- 
modulation products and harmonics to 
be created. The distortion products can 
be minimized with proper design of the 
amplifiers, while the harmonics are well 
attenuated with a low-pass filter at the 
output. The band switching is held to a 
minimum. 

The key to designing broadband 
amplifiers is feedback. Feedback can 
take a number of forms. Emitter degen- 
eration has been used in a number of 
designs throughout the book, and is one 
common form of feedback. Alone, how- 
ever, it is not sufficient in the design of 
broadband amplifiers. While it does have 
the effect of establishing a constant 
voltage gain where the output load 
resistance is established, it has the addi- 
tional effect of increasing the input 
impedance of the transistor. This in- 


crease is roughly proportional to the 
beta of the transistor. Since transistor 
beta is well approximated as fr+f in 
the high-frequency region, the increased 
beta at lower frequencies leads to an 
increasing input resistance as frequency 
is lowered. In a multistage amplifier, 
tli is leads to increasing gain with de- 
creasing frequency. 

One other form is shunt feedback. 
This usually takes the form of a resistor 
between the collector and the base of 
the transistor. This has two advantages. 
First, it stabilizes the current gain of the 
amplifier, an effect similar to the virtues 
of emitter degeneration. However, it 
also decreases the input and output 
resistances of the stages. 

Many examples have appeared where 
we have applied emitter degeneration 
alone. Shunt feedback may also be 
applied alone. Shown in Fig. 15 is an 
amplifier that uses shunt feedback. The 
transformer allows the 50-ohm load to 
appear as 200 ohms at the collector 
TTiis is adequate for a maximum power 
output on the order of 1/4 watt. The 
feedback path from the collector to the 
base contains a blocking capacitor, a 
small inductor and a 470-ohm resistor. 
The inductor has the effect of decreas- 
ing the feedback at high frequencies 
while the 470-ohm resistor is the domin- 
ant element at low frequencies. 

Measurements were performed with 
this amplifier. The transducer gain was 
measured in a 50-ohm system as 19 dB. 
The points where the gain was down by 
3 dB were 1 and 50 MHz. The upper 
limitation was the result of decreasing 
transistor gain the fy of the transistor 
was approximately 500 MHz. The low- 
frequency limit was a result of the 
transformer running out of inductive 
reactance. Only five bifilar turns were 


188 Chapter 8 









OPEN LOOP 















\ 


RE -10. Rf -250 


^3dB'V N 

\ 




N\ 


| 


rV 


100 


iooo 


FREQUENCY. MHz 


Fig. 17 - Transducer gain as a function of frequency for an amplifier with and without feed- 
back. The hybrid pi model of a bipolar transistor was used for this calculation. A dc beta of 

100 was assumed with f T = 500 MHz, C cb = 3 pF and V =50 ohms. Note that the gain with 
feedback is always lower than the open-loop gain (with no feedback) and that the bandwidth is 
always extended by application of negative feedback. 



Fig. 18 - C,, R in and R„ u , as a function of feedback components. A simple model was 
assumed for this calculation with a beta of 10. No account was taken for phase shifts in beta. 
Nonetheless, the calculations agree well with measured results. A profound advantage of feed- 
back is predictability in design. 


used on a small ferrite core (Amidon 
FT-2343). 

The transistor was biased for about 
120 mA of collector current. With this 
much current it would be expected that 
the output intercept might be fairly 
high. It was measured with two outputs 
of + 1 7 dBm each, or +23 dBm PEP (200 
mW) output. The intermodulation dis- 
tortion products were 40 dB down from 
each tone, indicating an output inter- 
cept of +37 dBm. The measurements 


were done at 10 MHz. 

Another approach to broadband 
design is to utilize a combination of 
emitter degeneration and resistive shunt 
feedback (see Fig. 16). This scheme has 
a number of advantages. First, it pro- 
vides two “handles” on controlling feed- 
back, which leads to greater llexibility. 
Second, the effect of feedback on im- 
pedance can be exploited. Since emitter 
degeneration has the effect of increasing 
input impedance, while shunt feedback 


decreases it, the combination effect 
causes the input impedance to be 
approximately constant. Also, the shunt 
feedback decreases the output imped- 
ance. leading to better interstage 
matching. Finally, emitter degeneration 
often has the effect of making an 
amplifier self-oscillate at some fre- 
quencies. This is especially true if the 
transistor has a very high f T . On the 
other hand, shunt resistive feedback 
almost always has the effect of making 
an amplifier unconditionally stable. This 
can be of significance in a high-gain 
amplifier chain. 

Shown in Fig. 17 is the effect of 
feedback upon transducer gain. This is a 
calculation based upon a transistor with 
a dc beta of 1 00, an fr of 500 MHz and 
a 3-pF collector-to-base capacitance. As 
shown, without feedback, die gain at 
low frequencies was over 32 dB. How- 
ever. the 3-dB bandwidth was only 8 
MHz. When a 10-ohm emitter resistor 
and a 250-ohm shunt feedback resis- 
tance were added, the gain dropped to a 
little over 10 dB. However, the 3-dB 
bandwidth is now extended to 65 MHz. 
The transistor model used in this anal- 
ysis is the so-called hybrid-zr model, 
and is covered in the appendix. 

If the amplifiers are to be cascaded, 
it is desirable that their input and 
output resistances be equal. Analysis 
shows that a rule of thumb may be 
applied. If the desired characteristic 
impedance is Z ot then the emitter resis- 
tance and the shunt feedback resistance 
should be chosen such that R e Rf = Z„ 2 . 

Fig. 18 has a curve showing the 
effect of emitter resistance upon stage 
gain, plus input and output resistance. 
The amplifier was designed for a 50- 
ohm characteristic resistance. Hence, for 
a given emitter resistance, R e< the feed- 
back resistor used was chosen according 
to the rule given above. That is, R f = 
(50) 2 ±R e . In this calculation, a simpler 
model was assumed for the transistor, 
with no account taken for a phase 
change of beta. The value of beta 
assumed was 10. In spite of the simple 
model, the results agree with the mea- 
surements we have done on amplifiers of 
this variety. It is interesting to note that 
the rule is a little away from the stated 
design center. That is, die input resis- 
tances are a little under 50 ohms, while 
the output resistances tend to be a little 
higher. Measurements confirm this cal- 
culation, also. 

The gain of a single stage may be 
increased over those values given in Fig. 
18 by the inclusion of a transformer in 
die output. The turns ratios are from 
1:1 to 4:1. It is not necessary that 
transmission-line transformers be used, 
although this may enhance performance 
in the vhf spectrum. 

Shown in Fig. 19 is a curve of gain 
vs. frequency for four different cases 


Modulation Methods 189 





100 kHz 1MHz 10 MHz 100MHz 10Hz 

FREQUENCY 


Fig. 19 — Transducer gain vs. frequency for amplifiers using transformers in the collector circuit. 
The hybrid pi model for a bipolar transistor was used in these calculations. The transistor speci- 
fications were the same as those in connection with Fig. 17. The low-frequency decrease in 
gain results from transformer characteristics. 



Fig. 20 - Example of a 30-dB gain broadband amplifier with 0.5 watt of PEP output (see textl. 


where transformers are used. The high- 
frequency rolloff is determined by 
transistor characteristics, while the low- 
frequency drop in gain is a result of the 
transformer model used. These calcula- 
tions were performed using tire hybrid-7r 
model which includes the effect of beta 
changes at high frequency, including a 
phase change. 

The information presented so far has 
dealt with small signal models. We have 
used the data to predict gain and input 
and output resistances for the ampli- 
fiers. Shown in Fig. 20 is a practical 
circuit where these ideas are applied. 
Assume that an output power of 1/2 
watt is desired from this amplifier. If 
this power is to be realized, the output 
load resistance presented to the collec- 
tor must be reasonably low. A 2:1 
transformer could not be used in the 
output, since this would place a 200- 
ohm load at the collector. This load 
would be too high unless a supply 
voltage greater than 12 were used. For 
simplicity, we will terminate this stage 
in 50 ohms, and ask for a gain of 1 0 dB. 
Looking at Fig. 18 we see that this level 
of gain can be achieved with an emitter 
resistance of 10 ohms and a shunt 
feedback resistance of 250 ohms. 

For this stage to deliver 1/2 watt of 
output, the dc input power must be at 
least one watt. In chapter 2 we found 
that the maximum efficiency which 
could be obtained from a Class A 
amplifier is 50 percent. Part of the 
supply voltage will be taken by a voltage 
drop across the 10-ohm emitter resis- 
tance. Hence, assume that the net 
supply available is 1 0 volts, to be placed 
across the transistor. This means that 
the current in the transistor will need to 
be at least 100 mA. To be on the safe 
side, we will bias it to 135 mA. Using 
the equation which relates output inter- 
cept to standing current in the transistor 
(presented in chapter 6 in connection 
with receiver front-end amplifiers), we 
would expect this circuit to have an 
output intercept of +40 to +43 dBm. If 
the number was +40 dBm, and the 
output power was 1/2-watt PEP (+27 
dBm), the output power in each tone 
would be +21 dBm, yielding IM prod- 
ucts that were 38 dB down. Such 
performance is reasonable to expect. 

Note that an rf choke is used to feed 
dc to the collector, and that another is 
used in the base-bias circuit. The choke 
is helpful in the latter case to prevent 
the input impedance of the output stage 
from being suppressed by the 100-ohm 
resistor in the bias divider. Also, since a 
500-ohm resistance is needed in the bias 
divider, but only 250 ohms were 
required for rf feedback, part of the bias 
divider is bypassed. 

Assume that a net gain of 30 dB was 
required from the amplifier. A 10-dB 
j^in is provided by the output stage. 


leaving 20 dB required from the driver. 
The output power required from the 
driver is only 50-mW PEP, or +11 dBm 
per tone in a two-tone test. If the IMD 
ratio for this stage alone is to be 40 dB, 
the output intercept should be +31 
dBm. An amplifier with a standing 
current of 50 mA should provide this 
performance. Looking at the curves, we 
see that the needed gain can be provided 
with a 2: 1-turns-ratio transformer in the 
collector, with a 5-ohm emitter resis- 
tance, and a 500-ohm feedback resistor. 


The other resistors in the circuit are 
chosen to provide the proper bias cur- 
rent for the transistor. 

An almost identical amplifier is de- 
scribed later as a construction project. 
The major difference is that the construc- 
tion project amplifier delivers 1-W PEP 
of sideband or 1 W of cw output. 

Various transistors may be used in 
amplifiers of this sort. Because of the 
heavy feedback employed, detailed tran- 
sistor characteristics are not of great 
importance. The f r of the devices 


190 Chapter 8 







should be at least 10 times greater than 
the highest frequency of operation. 
Also, the transistors should have suffi- 
cient power dissipation. Amplifiers of 
this kind are much different than the 
Class C amplifiers used for cw. The 
current in a Class A amplifier is con- 
stant, independent of the power output. 
Hence, the designer does not have the 
advantage of a low duty cycle that helps 
him when building cw rigs of similar 
power output. The writers have used the 
2N3553 for output stages at this power 
level in the hf bands. Although they 
have not been investigated experimen- 
tally, some of the transistors designed 
for the output of citizens band trans- 
ceivers should be ideal for these applica- 
tions. Devices worth consideration 
would be the Motorola MPS-U31 and 
MRF472. rhesc parts are relatively 
inexpensive. In any case, careful heat 
sinking is required because of the high 
power dissipation. 

If it is desired to extend the band- 
width of amplifiers of this kind to 
liigher frequencies, there are a few tricks 
that may be employed. From the curves 
it is evident that the widest bandwidths 
occur with the lower gain numbers. 
Because of this, a lower gain per stage 
will lead to increased bandwidth. 
Another trick that works well is to place 
a small inductor in series with the 
collector. This will increase the voltage 
swing on the collector at the upper 
frequencies while leaving the lower fre- 
quency gain unaltered. Values as low as 
50 to 100 nanohenrys are suitable for 
vhf work. Similarly, some inductance in 
series with the shunt feedback resistor 
will peak the high frequency gain. 
Finally, some impedance matching can 
be done. This would take the form of a 
pi or L type of network as an interstage 
coupling element. It should have a Q 
near unity, and should be tuned at the 
upper frequency of operation. It will 
then appear virtually “transparent” at 
the lower frequencies. 

It is sometimes desirable to run a 
Class A amplifier at even higher powers, 
although the power dissipations 
encountered *nay make the thermal 
designs difficult. Also, the high collector 
currents may make it difficult to use 
much emitter degeneration. This places 
the burden of feedback on the shunt 
element. Without a large emitter resis- 
tance, biasing will also be more cumber- 
some. A sample circuit is shown in Fig. 

2 1 . This amplifier is biased for a current 
of 1 A. With a 12. 5 -volt supply the 
input power will be a little over 10 
watts. The value of V cc is less than 12.5 
owing to die voltage drop across die 
collector resistor that is used as a 
current-sensing element for biasing. A 
2:l-turns-ratio transformer is used at 


Fig. 22 — Generalized schematic of a single-ended high-power Class AB rf amplifier. The rf the Output, transforming a 50-ohm ter- 

oicuitry is presented at A, while B emphasizes the details of the biasing circuit. mination to a 12.5-ohm load at the 


Modulation Methods 191 








collector. Because of the lack of emitter 
degeneration, the input resistance will 
be quite low. The base is matched with 
a composite 16:1 impedance ratio trans- 
former made from two “sortabaluns.” 
Although the writers have not built this 
amplifier, it should be capable of pro- 
viding about 5 watts of output through- 
out the hf spectrum with excellent IMD 
and high gain. For standby periods, or 
even keying, the circuit may be shut 
down by breaking the circuit at the 
point marked “X." It may be necessary 
to adjust R1 slightly to obtain 1 A of 
collector current. A large heat sink 
should be used at Ql. It would be 
advisable to provide some heat sinking 
at Q3 as well. 

High-Power Linear SSB 
Amplifiers The Biasing Problem 

For output powers exceeding 1 or 2 
watts, the Class A amplifiers outlined 
are not generally desired. The efficiency 
is too low, considering that the power 
must be dissipated on a continuous basis 
during the total transmit period. For the 
higher powers the more typical 
approach is to use a Class AB amplifier. 

Shown in Fig. 22 is a circuit for a 
typical linear amplifier for ssb service. 
No details are presented as to compo- 
nent value, for these will vary greatly 
with the frequencies of operation and 
the power levels desired. However, all of 
the circuits for this purpose follow the 
general form shown. 

In most ways the rf part of the 
design is exactly tire same as was pre- 
sented for cw amplifiers in chapters 3 
and 4. The output network should be 
designed for the peak-envelope output 
power and not the average power. That 
is, under two-tone testing conditions at 
a given I’HP level, the average power will 
be half the PEP. The output load 
presented to the collector is well 
approximated by R, = V cc 2 -r 2P oul . 
However, the power to use in the 
calculation is the PEP. If tire network 
were designed for average power, the 
amplifier would be voltage-limiting, 
leading to severe distortion of the flat- 
topping variety. 

The input resistance, input capaci- 
tance and output capacitance are well 
specified for most transistors designed 
for ssb power service. The networks are 
designed accordingly. The methods out- 
lined in earlier chapters may be used, 
with narrow -band or broadband trans- 
formers being suitable. 

The major difference between tire 
cw power amplifiers and the ssb ampli- 
fier is in tire biasing. If a cw amplifier 
were to be used for ssb service, severe 
distortion would occur. This would be 
most apparent at low levels. This is 
because the output transistor is cut off 
when there is no drive. The drive must 
be large enough to turn on the emitter- 


base junction to about 0.7 volt before 
any rf output occurs. 

The usual corrective method is the 
application of some forward bias. This 
establishes a quiescent operating current 
in the transistor when no rf drive is 
present. The base is already turned on, 
and the application of rf drive merely 
increases the base current. The dc 
collector current will increase accord- 
ingly. Unlike the case with Class A 
amplifiers, the transistor is not biased to 
full current on a dc basis. The level of 
quiescent current will depend upon the 
specific transistor used and is usually 
specified by the manufacturer. Values 
range from 15 to 1 00 mA. Probably the 
most informative reference is by Hejhall 
( QST for March, 1972 and Motorola 
AN-546). 

Fig. 22 shows a sketch of the usual 
biasing scheme used for this class of 
amplifier. The basis of the biasing is a 
diode: High-current type is the common 
choice. The transistor base bias should 
be chosen to deliver the desired quies- 
cent current in Ql under no-drive condi- 
tions. However, the bias should not vary 
more than 0.1 volt for all rf drive 
conditions. This means that tire dc 
current standing in the diode (supplied 
through R 1 ) should be larger than the 
peak current that will occur in the base 
of the transistor at times of maximum rf 
drive. 

The biasing of the amplifier is some- 
times aided by the resistance of the rf 
choke that isolates the bias diode from 
the rf energy at the base. This resistance 
allows a voltage divider action to occur 
which allows the bias diode to be run at 
a larger standing current than it would if 
the rf choke had no resistance. This 
extra current is used to supply base 
current during rf input peaks. The large 
bypass capacitor (500 juF) also helps to 
supply base current on a transient basis. 

The problems outlined here are com- 
plicated further by the thermal-runaway 
phenomenon. If Ql were in a virtually 
perfect thermal environment (a constant 
junction temperature), there would be 
no problem. This is not the case. When 
the transistor has rf drive applied for a 
period of time, tire resultant power 
dissipation will cause the junction tem- 
perature to increase. If the bias voltage 
is constant (as was advocated) tire 
higher temperature will cause the quies- 
cent current to be larger when drive is 
removed. If the increase is excessive, the 
collector current will be high enough 
that high-power dissipation will con- 
tinue within the transistor. This will 
lead to a further increase in junction 
temperature, causing an increase in qui- 
escent current. Thermal runaway is the 
ultimate result. 

There are solutions to this problem 
that are partially effective. One is to 
thermally bond the bias diode to the 


transistor. This causes the increase in 
transistor temperature to be communi- 
cated to the diode. Most silicon diodes 
which are fed from a constant-current 
source will show a voltage change with 
temperature of about -2 millivolt per 
degree C change. The negative sign 
indicates a decreasing voltage with 
increasing temperature, which is just the 
effect needed. 

Unfortunately, thermal bonding of 
the reference diode to the transistor is 
only partially effective. The reason is 
that the diode is capable of sensing only 
the temperature of the case of the 
transistor and not that of tire junction. 
The thermal resistance between the 
junction and the transistor case will 
allow the junction to run at a much 
higher temperature than that of the 
case. It is the junction temperature that 
controls current flow and ultimately 
leads to thermal runaway. 

One protective method is to include 
a diode within the transistor body for 
temperature sensing. The anode of the 
diode is brought out to a separate pin 
on the transistor and is used as a 
reference for a dc amplifier that pro- 
vides bias for the transistor. The reader 
is referred to the work of Chang and 
Locke (RCA note. AN-4591). 

The other technique is emitter 
degeneration. This can be external to 
the transistor. The more common situa- 
tion is where the degeneration is built 
into the device. Such transistors are 
referred to as containing “emitter 
ballasting." The advantage of the inter- 
nal degeneration is dial the emitter 
resistance may be distributed over the 
entire transistor structure with a separ- 
ate resistance element for each emitter 
section. The resistors are made from 
nichrome, which has a high -temperature 
coefficient of resistance. As a result, if a 
given section of the transistor begins to 
increase in temperature faster than 
others, that section is shut down faster. 
Such “hot spots" lead to second break- 
down, one of tire main phenomena that 
leads to destruction of power transis- 
tors. 

The experience of the writers sug- 
gests that transistors without internal 
ballasting must use external emitter 
degeneration if thermal runaway is to be 
avoided. This may not be vital in an 
amplifier to be used only for ssb service 
where the average power dissipation is 
low (because of the low duty cycle of 
tire human voice). However, if the 
amplifier is to be used for cw operation, 
or even if it is to undergo two-tone 
testing for linear service, some emitter 
degeneration must be used. Usually a 
fraction of an ohm will be sufficient to 
protect the transistor. 

In any case where emitter degenera- 
tion is used, either in tire form of 
ballasting or as external degeneration. 


192 Chapter 8 




Fig. 23 - Partial block diagram of an ssb transceiver. The system differs from an ssb transmitter in 
the inclusion of switching circuits and the multiple use of the carrier oscillator BFO and VFO. 


ihe resistance will cause the efficiency 
of the amplifier to be degraded. Also, it 
can have the effect of degrading the 
stability of the amplifier. Unconditional 
stability can sometimes be regained 
through the application of shunt feed 
back, at the cost of reduced stage gain 
Modern transistors designed for high 
power linear rf applications have excel 
lent 1MD specifications. Typically 


third-order distortion products are 30 
dB or greater below each tone during a 
two-tone IMD measurement at full out- 
put power. The distortion does not 
behave as nicely with such amplifiers as 
it does with a Class A design. With the 
Class A amplifier, an output intercept 
can usually be specified for a given 
circuit. This defines the IMD perfor- 
mance of ihe amplifier at all power 



Fig. 24 — A meihod for diode switching a crystal filter Only the input is switched in this 
example. A similar circuit could be used at the output. 


levels. Specifically, if the output power 
is decreased by X dB. the IMD ratio will 
improve by IX dB. Class AB amplifiers 
are not as well behaved. When the 
output power is dropped from the 
specified maximum, the IMD ratio can 
degrade. For this reason, the best mode 
of operation is at full rated power. If a 
low-level output is desired (for QRP 
experiments or driving vhf transverters), 
an attenuator should be used. Alterna- 
tively, the high-power final amplifier 
should be bypassed, with the output 
signal being obtained from an earlier 
Class A stage in the amplifier chain. 

One problem with the diode biasing 
scheme is the high current required to 
bias the diodes properly. This current is 
often obtained from the same power 
supply that is used for the collector 
bias. Most of the power used to derive 
the bias current is dissipated in the large 
resistor (Rl of Fig. 22). This will 
degrade the system efficiency consider- 
ably from that value given by the 
manufacturer. There are at least a 
couple of solutions to this problem. One 
would be to use a separate power supply 
for biasing the diode. The cost of a 
5-volt supply would be small. Another 
solution was suggested to the writers by 
W7UDM: Use the current that is stand- 
ing in a previous Class A amplifier to 
also bias the diode. The power is then 
used more effectively. Careful decou- 
pling would be required. 

No construction examples of high- 
power linear amplifiers are given in this 
chapter. However, some were presented 
earlier. They were designed for ssb 
service. 

Transceivers for SSB 

Although some operators use sepa- 
rate transmitters and receivers for ssb, 
the trend is toward transceivers. The 
major reason is convenience of opera- 
tion. With an ssb transceiver, once a 
station is tuned in so that it sounds 
proper in the receiver, the transmitter is 
automatically on the proper frequency. 
Another reason is that much of the 
transmitter and receiver circuitry can be 
shared, leading to economy in construc- 
tion. 

Shown in Fig. 23 is a partial block 
diagram of a single-conversion ssb trans- 
ceiver. The carrier oscillator used for ssb 
generation at the i-f is used also as the 
BFO for the receiver. It is not manda- 
tory that this signal be switched. It may 
be applied to both inputs simultane- 
ously. However, great care should be 
taken to ensure that minimal energy 
from the carrier oscillator finds its way 
into the receiver i-f amplifier. This 
avoids die noise-modulation problems 
which were reviewed in the receiver 
chapters. 

The VFO is also shared. Again, this 
signal may be applied to each of the 

Modulation Methods 193 





FROM O — \ 
RCVR • 




LOW +12V 

S?c 


Fig. 25 — Circuit for sharing a crystal filter between receive and transmit functions in a transceiver. Bipolar transistors are used at the input, while 
a dual-gate MOSFET is employed at the output. 


mixers simultaneously. If diode-ring 
mixers are used, it may be necessary to 
buffer each mixer input separately to 
ensure that proper LO injection levels 
are maintained. 

The third major component that is 
shared between the two functions is the 
crystal filter. It is usually necessary that 
switching be provided at at least one 
end of the filter, if not both. One 
solution would be to use diodes as the 
switching elements. A sample circuit is 
presented in Fig. 24. Only one side is 
shown, although the other side would 
be similar. Four diodes are used in this 
scheme. If input A is selected, CR1 will 
be conducting a dc current of about 25 
mA. CR2 is reverse biased by 6 volts. At 
the same time, the off channel (input B) 
is shunted to ground with CR3 which is 
conducting approximately 6 mA while 
additional isolation is provided by CR4 
which is back biased with 6 volts. The 
diodes may be 1N914 switching types 
for casual applications. However, better 
performance will probably be provided 
by using PIN diodes or low-speed high- 
voltage rectifier diodes. The reader is 
referred to the i-f amplifier discussion in 
chapter 5 for details. 

Another approach to filter switching 
is shown in Fig. 25. A pair of bipolar 
transistors is combined with a common 
collector connection to feed a 500-ohm 
crystal filter. The collector current in 
each transistor is determined by picking 
R3 and R4 appropriately. Small 200- 
ohm controls are used at R1 and R2 to 



IQjuF 

15V 

-±lf — O af input 

FROM SPEECH 
AMPLIFIER 


-±1( OAF OUTPUT TO 

10/lF RtcEIVER AFAMP. 

“isv 


Fig. 26 — Application of a diode ring as a balanced modulator during transmit periods, and a 
product detector for receiving. FETs are used for switching the audio. Q1 and Q2 may be 
general-purpose FETs such as the MPF102, 


194 Chapter 8 






establish the gain of each stage. The 
output of the filter is applied to a pi 
network consisting of Cl, C2 and LI. 
This network should be designed for a Q 
of 10 to 15, with resistances of 500 and 
2,700 ohms. The 2, 700-ohm resistor at 
the gate of Q3 ensures that the crystal 
filter has a proper termination. The 
output of the MOSFET amplifier is 
tuned to 9 MHz with a low-0 circuit. 
Two output links are used. One drives 
the receiver i-f amplifier while the other 
is applied to the transmit mixer. 

An innovative and unique means for 
ssb transceiver design is through the use 
of bidirectional circuits. These are cir- 
cuits that will function with signals 
flowing in either direction. One example 
that has been discussed in detail is the 
diode-ring mixer. An example is shown 
in Fig. 26 with a circuit that functions 
both as a balanced modulator during 
transmit periods and as a receiving 
product detector. JFETs are used as 
switches at the audio port. Point "A" 
should be high (+12 volts) during trans- 
mit periods and point “B” positive for 
receiver operation. 

Shown in Fig. 27 is an amplifier that 
will provide gain in either direction. The 
direction is controlled by choosing 
which power-supply port is activated 
with +12 volts. Each transistor is biased 
for a current of approximately 35 mA, 
which is enough to yield good 1MD 
performance. A 2:I-turns-ratio ferrite 
transformer is used in the output of 
each collector in order to obtain some 
impedance matching. However, this 
could be eliminated if lower gain is 
desired. Provision is made for the use of 
both shunt and series (emitter) feed- 
back. Again, depending upon the gain 
desired, one or the other may be elimin- 
ated. Some shunt feedback would be 
desirable in order to preserve stability, 
since the transistors specified have an f T 
of over 1 GHz. It is important that the 
two stages share a common emitter 
resistance as part of the dc biasing 
scheme. This will ensure that the “off 
transistor has both of its junctions 
reverse biased. This circuit is an adapta- 
tion of one designed by W7UDM. 

Bidirectional circuits are ideal for 
driving the rf and i-f ports of a diode- 
ring mixer. When used in this way, the 
only switching required would be that 
for controlling the direction of the 
amplifiers. 

Shown in Fig. 28 is a partial block 
diagram of a possible ssb transceiver 
that could be built with diode-ring 
mixers and bidirectional amplifiers. If 
desired, a PIN diode attenuator or two 
could be inserted in the signal path for 
control of gain in both modes. Tech- 
niques of this kind have been used in a 
commercially built multiple-conversion 
transceiver. However, the amplifiers 
used germanium transistors biased for 



Fig. 27 - Circuit for a bidirectional amplifier using bipolar transistors. Q1 and Q2 are 2N5943 
or similar devices with a high fy. 



Fig. 28 - Partial block diagram of an ssb transceiver based upon bidirectional circuits. 


minimal current. Many of the mixers Double-Sideband Transmitters 
were designed similarly. The dynamic In the earlier theoretical discussion 
range ot the system was disastrous! On we treated suppressed-carrier double 
the ot her hand, using these concepts a side band as a intermediate step toward 
good 20-meter ssb transceiver has been generation of an ssb signal. While this is 

designed and built by W7UDM. By using normally the case, there are many situa- 
. diode-ring mixers with proper LO injec- tions where a double-sideband trans- 

tion and amplifiers with adequate cur- mitter is quite useful. An advantage of 

rent, and by using single conversion, a dsb over ssb is simplicity. The major 

receiver dynamic range ot 90 dB has disadvantage is that extra spectrum is 

been obtained. The advantage of this occupied. Sometimes, the tradeoff may 

scheme is that virtually all of the filter- favor the use of dsb. 

ing in the system can be used for both One application for dsb that comes 
transmit and receive. This is highly to mind is for the QRP enthusiast, 

desired. In any transceive system, it is Often he has an interest in working 

advisable to run the received signal phone, but has little interest in building 

through the low-pass filters that will be a complete ssb transmitter. Dsb gives 

needed for harmonic filtering ot the him an alternative. Another point in his 

power amplifiers. favor is that transceivers are built easily 


Modulation Methods 195 





to utilize the VFO which is already 
present in a direct-conversion receiver. 
All of the normal advantages of a ssb 
transceiver (in contrast to a separate 
transmitter) are available. Specifically, 
once a station is tuned in with the 
receiver, the transmitter is automatically 
on the same frequency. There is an 
additional advantage: If an unused fre- 
quency is found with a direct- 
conversion receiver, the user can be 
assured that the segments on either side 
of his carrier frequency are unoccupied. 
If he were to call CQ, he would not be 
causing undue interference as a result of 
his extra sideband. Additionally, if an 
ssb station is copyable with a “dc” 
receiver, the operator knows that he 
may call that station without causing 
QRM to an adjacent channel. If that 
channel were occupied, it would have 
been heard in the direct-conversion 
receiver. 

There is a liability with the trans- 
ceiver using a “dc” receiver and dsb 
transmitter. Two such units are not 
compatible with each other. A dsb 
station is not generally copyable with a 
dc receiver. This is normally not a 
problem for the QRP operator, for most 
of his contacts are with higher-power 


ssb stations. Because of the spectrum 
used, dsb is not recommended for use 
on the hf bands except at low powers. A 
maximum limit might be 10-watts PEP 
output. 

Another application for the dsb 
transmitter would be for the DX- 
oriented vhf operator. He often has a 
desire to converse with local ssb opera- 
tors with common interests. For such 
purposes low power is usually sufficient. 
When band openings occur and the 
more distant contacts are available, he 
switches to cw to ensure the contact. 

The vhf dsb station again has the 
liability that half of his transmitted 
power is in an unwanted sideband. On 
the portions of the vhf bands where ssb 
and cw predominate, the extra spectrum 
space occupied by dsb is rarely a prob- 
lem. The mountain-topping vhfer might 
consider it wasteful to throw away 3 dR 
of extra energy from his battery pack. 
However, if he were to examine the 
current that would be required to 
remove the extra sideband, the differ- 
ence becomes much less significant. This 
is especially true for the portable station 
running less than 1 watt of output. 

A final advantage of building a dsb 
vhf transmitter is that it is expandable. 


The oscillator (and multiplier chain, if 
used) can always be adapted for use 
with a later ssb exciter. The balanced 
modulator and speech amplifier may 
also be used later with some modifica- 
tion to another frequency. A linear 
amplifier chain designed specifically for 
a dsb transmitter may be used directly 
with a later ssb replacement. 

A Simple DSB Transmitter 
for Six Meters 

Shown in Fig. 29 is a simple QRP 
transmitter for 6 meters. A third- 
overtone crystal oscillator is operated 
directly at the output frequency. This 
circuit delivers about +10 dBm of drive 
to the balanced modulator. The bal- 
anced modulator is simple, using two 
hot-carrier diodes driven from a ferrite 
transformer. This circuit uses a pair of 
variable capacitors for adjustment of the 
carrier balance. Over 50 dB of carrier 
rejection was measured with this cir- 
cuit on an open bench when driven 
from a separate signal generator. In the 
transmitter shown the carrier suppres- 
sion is less - only 36 dB. This is because 
the signal from the crystal oscillator is 
leaking around the balanced modulator 
to the amplifier chain. Some additional 



Fig. 29 - Circuit for a 6-Meter dsb QRP transmitter (see text). T-R switching is realized with a double-pole, double-throw slide switch. 

LI — 10 turns No. 24 enameled wire on L3 — 6 turns No. 22 enameled wire on T1 — 10 trifilar turns No. 30 enameled wire 

Amidon T37-6 toroid core. Amidon T50-6 toroid core. on Amidon FT-37-61 toroid core. 

L2 - 2-turn link over LI. 


196 Chapter 8 









Interior view of the 60-MHz dsb transmitter. The crystal oscillator, speech amplifier and 
balanced modulator are on this circuit board. 


isolation would solve this problem. 

The speech amplifier consists of a 
single 741 operational amplifier. The 
feedback resistors were picked to pro- 
duce a suitable output level while using 
a microphone from an inexpensive 
cassette tape recorder. 

A test point is provided in the 
balanced modulator. If +12 volts are 
applied to this resistor, the circuit is 
unbalanced, yielding a carrier output for 
test purposes and alignment. If the 
transmitter is to be used on cw, this 
point could be keyed to the + 12-volt 
supply with a pnp switch. In these 
applications, it would be wise to also 
key the supply to the linear-amplifier 
chain. 

The linear amplifier uses four stages 
with an output of 400-mW PEP. The 
first three stages were designed for 10 
dB of gain per stage, with heavy nega- 
tive feedback being employed in each 
stage. The output has shunt feedback, 
but there is no emitter degeneration. 
Because of this the gain is not as flat 
with frequency as it is in the preceding 
stages. A 6-dB attenuator is used at the 
input to the amplifier chain to ensure 
that a proper termination exists for the 
balanced modulator. 

The first two stages in the amplifier 
chain use 2N5179 transistors. These 
devices have an / T of 1 GHz and a low 
collector-to-base capacitance. They are 
recommended for general-purpose vhf 
use. The driver and output amplifier use 
Arnprex A-2 1 Os. This transistor is 
nigged and has an fo of 1200 MHz. A 
suitable substitute would probably be 
the 2N3866 or the 2N3S53. Since the 
standing current is moderatly high (over 
100 mA in Q5), heat sinks are needed 
for Q4 and Q5. 

A small piece of double-sided pc 
board was used for the crystal oscillator 
and the balanced modulator. The top 
side, where the components reside, was 



External view of the 50-MHz dsb transmitter. 
The slide type T-R switch is adjacent to the 
BNC connectors which are used for the 
antenna and the line to the receiver. Power 
receptacles are also close to this switch. The 
crystal socket and microphone jack are at the 
opposite grid of the chassis. 



The driver is the transistor with the small 
heat sink. A slightly larger heat sink is used 
on the output amplifier, which is hidden 
below the small board containing the output 
network. 


used as a ground plane. The amplifier 
chain was built on single-sided board. 
The extensive use of feedback makes 
ground-loop problems less severe. The 
board was originally etched as a general- 
purpose instrumentation amplifier 
(described in chapter 7) which dictated 
the circuit configuration. If higher-gain 
circuits were used, employing 2:l-turns- 
ratio transformers in the outputs of the 
low-level stages, it would be possible to 
obtain the needed gain with three 
stages. The present amplifier has a small' 
signal gain of 45 dB at 50 MHz. 

It should be straightforward to adapt 
this circuit to any of the lower bands. 
The bypass capacitor at the emitter of 
Q5 should be removed in order to drop 


the gain accordingly. The crystal oscilla- 
tor and the simple pi network output 
would be replaced with suitable circuits. 
If the amplifier is to be used on the 
160-meter band, it would be advisable 
to increase the inductance value of the 
rf chokes to around 50 /all. The output 
network is adjusted for maximum out- 
put with the test point set to +12 volts. 
The output should be monitored in a 
high-frequency oscilloscope for fiat 
topping (if such an instrument is availa- 
ble). Good results have been obtained 
with this transmitter. 

A DSB/CW Exciter for 144 MHz 

Experience with the 6-meter QRP 
dsb transmitter was encouraging. A simi- 
lar unit was built for the 2-meter band. 
A number of refinements were included 
for operational convenience and to test 
a number of experimental ideas. The 
circuit for the transmitter is shown in 
Fig. 30. 

While crystal control is adequate for 
some operations, flexibility in fre- 
quency coverage is highly desirable. 
There are a number of ways to achieve 
this at vhf. The usual one is to use a 
heterodyne type of transmitter circuit. 

An alternative to a heterodyne 
exciter is to use a low-frequency VXO 
and a multiplier chain. While a VFO 
could have been used, it is quite diffi- 
cult to obtain suitability for cw and ssb 
at vhf. A Colpitts crystal oscillator was 
modified with an inductor and variable 
capacitor in series with the crystal. With 
this circuit (Ql), approximately 100 
kHz of tuning range in the 2 -meter band 
was obtained. The frequency shift could 
have been extended farther. (See VXOs 
in chapter 2.) 

The frequency-multiplier chain was 
unconventional, but highly successful. A 
frequency of 18 MHz was chosen for 
the VXO, allowing the 2 -me ter band to 
be reached by using frequency doublers. 
The output of the oscillator is buffered 
and filtered in order to yield a 
symmetrical waveform with a power of 
over +10 dBm. This output was then 
applied to a balanced doubler which 
uses a pair of silicon switching diodes. 
The output of the doubler was filtered 
in a single tuned circuit, furnishing 
energy at 36 MHz. This was amplified to 
a + 10-dBm level with a broadband 
amplifier. The same methods were 
repeated to arrive at 72 and finally 144 
MHz. The 144-MHz output was filtered 
with a double-tuned circuit, providing 
power output of + 1 1 dBm. 

The output of the multiplier chain 
was carefully investigated with a spec- 
trum analyzer to evaluate the spurious 
responses. Only one spur could be 
found. That was at 72 MHz. It was 55 
dB down. All other subharmonic spurs 
were undetectable. This response is a 
result of using simple balanced circuitry 


Modulation Methods 197 



Fig. 30 - Circuit diagram for a 144-MHz cw/dsb transmitter. See text for details. Variable capacitors are air. Teflon, or ceramic- 
dielectric types. All resistors are 5 percent. 1/4 watt. 

Cl — 5-80 pF air variable. a T37-6 toroid core. L4 — Air core. 0.25 ID X 0.65 long finch) 

LI 24 turns of No, 27 enameled wire on L3 — 12 turns of No. 27 enameled wire on 10 turns of No. 22 enameled wire, taps 

a T37-6 toroid core. a T37-6 toroid core, 3-turn input link, at 1-1/4 and 1-3/4 turns. L7 

L2 — 14 turns of No. 27 enameled wire on 2-turn output link. L5 — 5 turns, air core, 1/4 ID X 1/2 L® 

198 Chapter 8 






+ 12V 39 


+ 12V 

AMPLIFIER 

.01 



EXCEPT AS INDICATED, DECIMAL 
VALUES OF CAPACITANCE ARE 
IN MICROFARADS ( jiF) ; OTHERS 
ARE IN PICOFARADS ( pF OR jijiF); 
RESISTANCES ARE IN OHMS', 
k -1000. M. I 000 000 


+ 11dBm 




100k I 1000 . ,7T3^r 


X^ 



BNC TO 
ANTENNA 1 


long (inch), taps at 1 and 3/4 turns. 
L6 — 5 turns, air core, taps at 1 turn 
and 2-1/2 turns. 

L7 — 5 turns, air core, tap at 1 turn. 
L8 - 10 turns No. 27 enameled wire on a 


T37-6 toroid core. 

L9 - 7 turns, aircore, taps at 3/4 and 3 turns. 
T1.T3, T5. T7 - 7 trifilar turns No. 30 
enameled wire on an FT-23-43. 


ferrite core. 

T2, T4, T6, T8, T9 - 5 bifilar turns No. 30 
enameled wire on an FT-23-43 
(ferrite) core. 

Modulation Methods 199 





Interior view of the 2-meter transmitter 
showing the oscillator and multiplier chain. 
The lower board contains the 18-MHz VXO, 
a buffer, and the first diode-doubler ampli- 
fier combination. The upper board contains 
two more diode-doubler amplifier combina- 
tions and a double-tuned 144-MHz output 
network. In spite of the open construction, 
the output of the chain is remarkably free of 
spurious responses. 



This board contains the balanced modulator and rf power-amplifier chain for the 2-meter 
exciter as well as keying circuits and the speech amplifier. The stud of the 2N5947 output 
amplifier is attached to a small piece of aluminum which serves as a heat sink. 


rather than relying upon shielding or 
selectivity. It was found that hot carrier 
diodes gave superior performance in the 
last frequency doubler. While the out- 
put power was sufficient with lN914s, 
the 72-MHz component was only 50 dB 
below the desired output. 

Other frequency-multiplier schemes 
were investigated. While single-ended 
multipliers were the simplest, double- 
tuned circuits were required at each 
frequency in order to keep spurs 50 dB 
down. Push-push doublers were tried 
using well-matched transistors. While 
the suppression of fundamental drive 
was good, instability problems were 
encountered in cascading a number of 
such stages. The diode frequency 
doublers have been found to be one of 
the best avenues to follow for frequency 
multiplication. The broadband ampli- 
fiers appear to be unconditionally stable 
and the tuning is unambiguous. A more 
exotic filter at die output (L6 and L7) 
would suppress the spurs by even higher 
ratios. 

The output of the frequency- 
multiplier chain is applied to a balanced 
modulator to generate the dsb signal 
directly at 144 MHz. The balanced 
modulator and speech amplifier are 
virtually identical to those used in the 
50-MHz transmitter. The differences are 
a reduced number of turns on a smaller 
ferrite core and the use of smaller 
balancing variable capacitors. The trans- 
mitter strip was originally built and 
adjusted in the home shop. As adjusted, 
tlie carrier suppression was 40 dB. When 
it was adjusted more carefully while 
using a spectrum analyzer, a suppression 
of over 50 dB was obtained. Using an 
outboard signal source (+13 dBm), simi- 
lar levels were obtained at 14, 28 and 
50 MHz. “Retweaking” was required at 
each band. The carrier suppression was 
only 35 dB at 220 MHz. 

The linear-amplifier chain is similar 


to that shown for 50 MHz (Fig. 29), 
although only three stages were used. 
The input stage, Q6, used a 2N5179 
while die driver, Q7, used a 2N3866. 
Bodi of these stages were designed for 
20 dB of low-frequency gain per stage, 
and included a ferrite transformer in the 
collector circuits for matching. The out- 
put amplifier, Q8, used a Motorola 
2N5947. This stud-mount transistor is 
specified for Class A linear service. The 
stage was set for a gain of near 10 dB 
with a collector current of 120 mA. The 
collector rf choke is a toroidal inductor. 
A piece of aluminum with an area of 
five square inches served as a heat sink 
for Q8. The weakest link in die trans- 
mitter is the output network which used 
a single tuned circuit. The taps were 
adjusted for maximum cw output power 
while using home-lab type equipment. 
Later measurements revealed that the 
second harmonic at 288 MHz was only 
suppressed 20 dB. This presented no 
problems in operation, since an out- 
board filter was used. An improved 
output network is definitely in order 
and should certainly not be difficult. An 
L-C-L type of T network should provide 
suitable performance, as would a double 
pi circuit. 

The balanced modulator, 6-dB pad 
and output network were disconnected 


and the amplifier was evaluated over a 
wide frequency range. The gain at 144 
MHz was 37 dB, while 48 dB was 
available at 50 MHz. The gain at 220 
MHz was down to 31 dB. Some induc- 
tance in the collectors of the three 
stages would peak this up if operation 
on that band was contemplated. Alter- 
natively, another low-level 2N5179 
amplifier could be used. The gain at 28 
MHz was nearly 50 dB. However, at 
lower frequencies the gain began to 
drop. This is predominantly because of 
the 470-pF coupling capacitors used. 
The output power was 400-mW PEP dsb 
or cw at 144 MHz. 

Cw operation is provided by keying 
the + 12-volt supply to the total ampli- 
fier chain. The dc that is applied to the 
balanced modulator was also keyed. The 
backwave of this transmitter was mea- 
sured at —75 dB. An RC network is 
included for shaped keying. 

The construction method used for 
this rig was unorthodox for vhf. A large 
piece of double-sided pc board was 
etched to form some breadboard mate- 
rial. The top side was a matrix of copper 
islands, 1 cm on a side. The back of the 
board was a continuous ground foil. The 
same results can be achieved with a 
hacksaw. The capacitances of the board 
presented no problems because almost 



Exterior view of the 2-meter dsb/cw transmitter. The knob controls the frequency of the VXO 
at 18 MHz. 


200 Chapter 8 




Interior of the 75-meter transceiver. The VFO compartment is at the bottom of the photo- 
graph, and the receiver board is seen at the center. The transmitter output amplifier, 
balanced modulator and speech amplifier are mounted on the end panel at the top of the 
picture. 


all of the high-frequency circuitry was 
at a low impedance level. The capaci- 
tance of each pad section was less than 
0.5 pF. Holes may be drilled through to 
the ground foil wherever a ground 
connection is needed. The VXO and 
first doubler are on one board. A second 
board contains the other two frequency 
doublers. A third board contains the 
balanced modulator, speech amplifier 
and linear-amplifier chain. Results with 
this transmitter have been good. 

A 75-Meter Transceiver — 
Direct-Conversion Receive 
and DSB Transmit 

The transceiver described in this 
section covers the 80-meter cw and 
75 -me ter phone bands. It provides full 
transceive and has an output of over I 
watt. This rig was built by Jeff Damm, 
WA7MLH, and is used for home station 
and portable operation. 

Tire VFO section of the transceiver 
is shown in Fig. 31. This circuit is 
similar to many that have been used in 
other projects. The Hartley configura- 
tion is used. Reasonable stability is 
obtained by using capacitors of both the 
NPO ceramic type and air variables. An 
MPF102 JFET is used and is Zener- 
diode regulated. The VFO is tuned with 
a capacitor from a surplus BC454 re- 
ceiver. This capacitor has a maximum 
range of nearly 200 pF. The VFO 
requires that the variable capacitor (in 
parallel with the inductor) cover a range 
of 33 to 68 pF in order to tune the 
range from 3.5 to 4 MHz. In the 
WA7MLH transceiver a combination of 
fixed-value ceramic NPO and air-variable 
capacitors was used in series with the 
main tuning capacitor to obtain the 
proper range. 

The VFO is built in a separate box 
that is contained within the main cab- 
inet. Since the oscillator operates at the 
same frequency as the transmitter out- 
put, it is important that good isolation 
be maintained. The oscillator is buffered 
with a feedback amplifier consisting of 
Q2 and Q3. The output power available 
is +10 dBm into a 50-ohm termination. 
The emitter current in Q3 was chosen 
large enough to maintain a sine-wave 
output under a 50-ohm load. 



External view of the 75-meter dib/ew trans- 
ceiver built by WA7MLH. The VFO control 
is at the left. 


In transmit the VFO output is 
applied to the balanced modulator 
shown in Fig. 32, using a two-diode 
circuit. Carrier balance is adjustable 
with a 100-ohm control between the 
diodes. The carrier suppression was 36 
dB. The IN914 diodes were matched 
for forward resistance by means of an 
ohmmeter. 

The output of the balanced modula- 
tor is applied to a 6-dB pad to assure 
proper termination, and is then routed 
to a broadband amplifier, Q4. This stage 


provides nearly 20 dB of gain. The 
balanced modulator and the first linear 
amplifier (Q4) are contained on a single 
circuit board. 

Another circuit board contains the 
speech amplifier and a pnp transistor, 
05, for cw keying. The speech amplifier 
uses a pair of 741 op amps. Keying is 
realized through addition of Q5, a 
2N3906. 

The output amplifier is shown in 
Fig. 33. A 2N5189 transistor is biased 
for a current of 50 mA and serves as the 



Fig. 31 - 3.5- to 4-MHz VFO for the WA7MLH 75-meter dsb transceiver. 

Cl — 200 pF. Air variable capacitor. Amidon T68-2 toroid core, tapped 12 

LI - 51 turns No. 26 enameled wire on turns from ground. 


Modulation Methods 201 





Fig. 32 — Balanced modulator, speech amplifier and keying switch for the WA7MLH transceiver. 
T1 — 15 trifilar turns No. 30 enameled wire T2 — 12 bifilar turns No. 30 enameled 
on an Amidon FT-37-61 toroid core. wire on an FT-37-61 toroid core. 


with a dual-gate MOSFET product 
detector. While sensitivity was more 
than sufficient, severe problems were 
encountered with square-law detection 
of a-m stations. The MC1496G detector 
eliminated these problems with no pen- 
alty in sensitivity. 

Transmit-receive control is provided 
by SI, a double-pole, double-throw 
toggle switch. One set of contacts 
switches the 12-volt supply between the 
transmitter and the receiver. A 12-volt 
relay is controlled by this line to change 
the antenna from the receiver to the 
transmitter input. The other contacts on 
SI disconnect the output of tire speech 
amplifier from the balanced modulator 
during receive periods. Without this 
measure, the operator’s voice could be 
heard in the receiver during that mode. 
The + 12-volt supply should be applied 
to the speech amplifier continuously. 

A useful addition to this transceiver 
would be a meter (0-1 A) to monitor 
the total power-supply current. The 
operator could then adjust his voice 
level and microphone gain such that the 
current remained constant during trans- 
mit periods. An increase in current 
would indicate that the final amplifier 
was being overdriven. This would in- 
crease the distortion products signifi- 
cantly. Excessive “fiat topping" was 
observed on an oscilloscope when the 
linear amplifier was overdriven. 

A Universal Exciter for SSB and CW 

The transmitter described in this 
section was designed to provide good 


driver. A small heat sink is used on this 
stage. The output amplifier, Q7, is an 
inexpensive plastic power device, a GE 
type D44C6. It is biased for Class A 
operation with a standing dc collector 
current of 250 mA. The saturated cw 
output power of this stage is 1 .5 watt. 

About I -watt PEP of dsb output power 
is obtained. A half-wave filter serves as 
the output network. 

Measurements have been performed 
on a similar breadboarded version of 
this amplifier. The overall gain of the 
linear chain (Q4, Q6 and Q7)is over 40 
dB. The same gain is available in the 
40-meter band. The gain drops signifi- 
cantly at 14 MHz. This results from the 
limited fr of the output transistor. If a 
similar transmitter were to be used on 
the higher amateur bands, an output 
transistor with a higher f T would be 
desirable. 

The receiver used in the WA7MLH 
transceiver uses an MC1496G product 
detector which is followed by a pair of 
audio amplifiers containing 2N3565s. 

This receiver was described in detail in 
chapter 5. The original version of the Fig. 33 - Rf-output amplifier and details of T-R switching for the WA7MLH dsb transceiver. 
WA7MLH transceiver used a receiver System is shown in the transmit position. 



202 Chapter 8 





OSCILLATOR 


BALANCED MODULATOR 


^OOO 1000, 


100 * 


Fig. 34 - Carrier oscillator, speech amplifier and balanced modulator for the universal ssb transmitter. Insert shows the FET oscillator used in 
the KL7IAK version. Coils are identical for either circuit. A double-pole, double-throw switch (SI ) serves as the mode switch. Any type is 
suitable since no rf is switched. The other half of the switch is shown in Fig. 41. All variable capacitors are mica compression or ceramic 
trimmer types. 

LI — 45 turns No. 28 enameled wire on L2 — 3 turn link over LI . wire on an Amidon T50-6 toroid core, 

an Amidon T50-2 toroid core. L3 - 20 bifilar turns No. 28 enameled L4 - 6-turn link over L3. 


performance on ssb and cw. It was 
intended primarily for QRP work. The 
output power is enough that higher 
power linear amplifiers may be driven 
directly. Data are given for operation on 
any amateur band from 1 .8 to 50 MHz. 
The original unit was built by Terry 
White, KL7IAK. 

The transmitter was a single-band 
unit for 20 meters. The filter approach 
to side-band selection was used and a 
narrow -band design was adopted for the 
rf ptwer chain. 

I Shown in Fig. 34 is the carrier 
generator, balanced modulator and 


speech amplifier. The carrier oscillator 
uses a pair of bipolar transistors. A 
common tuned circuit is shared by the 
collectors of the two oscillators. How- 
ever, only one transistor is biased “on" 
at a time. This allows the operator to 
choose the desired sideband. The JFET 
oscillator used for usb generation in the 
original KL7IAK unit is also shown in 
the insert in Fig. 34. 

An MC1496G is used as a balanced 
modulator. Means are provided for 
adjusting the carrier balance. Measured 
carrier suppression was over 50 dB. 
Code operation is realized by inserting 


dc into the balanced modulator. This 
allows sufficient carrier energy to ride 
through for cw operation. 

The speech amplifier uses a JFET 
input amplifier, making the circuit com- 
patible with high- or low-impedance 
microphones. The FET is followed by a 
741 op amp which provides a voltage 
gain of 10. If additional gain is needed, 
a second op amp could be cascaded with 
die first. 

Shown in Fig. 35 is the i-f and 
output mixer system for the trans- 
mitter. A pair of dual-gate MOSFETs is 
used as 9 -MHz amplifiers. They provide 


Modulation Methods 203 














Fig. 35 - l-f amplifier and transmit mixer for the universal ssb transmitter. The insert shows the mixer output circuit used in the KL7IAK 
version of this exciter. R1 is a pc-board-mounted control. 

L5 — 28 turns No. enameled wire on an 
Amidon T50-6 toroid core. 

L6 — 3-turn link over L5. 


T1 — 15 bifilar turns of No. 30 enameled 
wire on an FT-37-43 (primary). 5-turn 
secondary. 


T1 A — see text. 

Z1 — 9-MHz crystal filter, KVG type 
XF-9A. 


some signal gain, terminate the crystal 
filter, and provide a convenient means 
for adjusting the gain. The application 
of gain control to gate 2 of a dual-gate 
MOSFFT amplifier was discussed in the 
receiver chapters. While this can cause 
IMD to be generated, the signal levels in 
this i-f amplifier are low enough that it 
is not a problem. If desired, an ale signal 
could be applied to the two gates. The 
reader is referred to the receiver 
chapters for the discussion of age 
systems. 

The output mixer transfers the 
9-MHz ssb signal to the output fre- 
quency of interest. An MC1496G is 
used as the mixer. The 1C is biased for 
larger currents than are normally used 
with this device. This enhances the 
linearity (the output intercept is 
increased). Broadband and narrow-band 
output networks are shown. The broad- 
band transformer will provide a 50-ohm 
output over a wide range of frequencies, 
making it suitable for driving multi- 



The 9-MHz i-f amplifier used in the universal 
ssb/cw transmitter. 


section filters at the desired output 
frequencies. See Fig. 38. The alternative 
narrow-band output (used in the 
KL7IAK version) uses a tuned trans- 
former. For 14 MHz, the primary has 20 
bifilar turns of No. 30 wire on an 
Amidon T50-6 core. The secondary has 
a 3-turn output link. The narrow-band 
transformer has enough bandwidth to 
cover the entire 20-meter band, but still 
offers some image rejection. The 
narrow-band output is suitable for 
adaption to most of the hf bands. For 
use on 160 or 75 meters, it would be 


desirable to use the wide-band design. 

Shown in Fig. 36 is the circuit for 
the 5- to 5.5-MHz VFO that is used in 
the KL7IAK 14-MHz version. The 
reader is referred to the 80- and 20- 
meter superhet receiver in chapter 5, 
and to the discussion of VFOs in 
chapter 3. The VFO should be capable 
of delivering a signal to the MC1496G 
mixer of 1 volt pk-pk across 50 ohms 
(+4 dBm). 

The narrow-band linear rf-amplifier 
chain used in the K.L71AK transmitter is 
shown in Fig. 37. This circuit uses three 



Fig. 36 - A 5.0 to 5.5-MHz VFO for the universal ssb transmitter. This circuit may be used as 
shown for 3.5 to 4, or 1 4 to 1 4,5-MHz operation. For other bands it is heterodyned to the 
appropriate injection frequency. This is presented in Fig. 40. L7 is a 3.4-pH inductor on a 3'~ 
inch diameter ceramic form (no tuning slug used). Cl is a 150-flF air variable. 


204 Chapter 8 










stages and delivers an output of 2.5- 
watts PEP, or cw with a total small- 
signal gain of 57 dB. The input stage is a 
2N5859 biased for a current of 25 mA. 
This is followed by another 2N5859 
which runs at a collector current of 60 
mA. A tuned transformer is used in the 
output of the input stage. A pi network 
matches the driver to 50 ohms. The first 
two stages are capable of delivering 100 
mW of output with excellent linearity. 
This stage was matched to 50 ohms 
rather than directly to the base of the 
final amplifier to allow the low-level 
output to be extracted for driving vhf 
transverters. 

The output amplifier contains a 
Motorola 2N6366. The networks were 
designed from the impedance data 
supplied by the manufacturer. A C-C-L 
type of T network is used for base 
matching and a pi network was em- 
ployed for the output. Originally, the 
circuit was built following the sample 
presented in the manufacturer's litera- 
ture. The transistor was bolted to a heat 
sink and the reference diode was 
soldered to a lug that was fastened to 
the stud of the transistor. The perfor- 
mance appeared to be exactly that 
specified by the manufacturer when R1 
was set for an idling current of 15 mA. 
However, the amplifier could be run 
only for very short periods in a cw or 
two-tone ssb test. If the operating 
period exceeded half a minute, the 
transistor would go into thermal run- 
away. If rf drive was applied for 1 
minute, then removed, the collector 
current was near 200 mA. At this level 
tire heating was enough without rf drive 
that current would slowly increase. 

In order to ensure thermal stability, 
emitter degeneration was inserted in the 
circuit. Four 2.7-ohm, 1 /4-watt resistors 
were paralleled to provide a resistance 
of 0.68 ohm. The bias in the diode was 
then readjusted for 15-mA collector 
current with no rf drive. The stage ^in 
was decreased by means of the emitter 
degeneration. A slight instability was 
cured by placing a 220-ohm resistor 
across the collector rf choke. 



Front view of the 20-meter version of the 
universal ssb/cw transmitter (built bv 
KL7IAK). 



The 20-meter power-amplifier chain used in the universal ssb/cw transmitter. The input stage is 
seen at the left. To the right is the Class AB output amplifier. Power output is approximately 
2-1/2 W cw or PEP. Third-order IMD products are 30 dB below the 2-W PEP two-tone output 
with this amplifier. 



Fig. 37 - A 14-MHz narrow-band rf-power amplifier chain used in the KL7IAK version of the 
universal ssb transmitter. All variable capacitors are mica-compression types. R2 is 0.68 ohm 
(four 2.7 ohm, 1/4-watt, 5-percent resistors in parallel). 

L8 - 18 turns No. 22 enameled wire on a L10 - 9 turns No. 22 enameled wire on a 

T50-6 toroid core. T50-6 core. 

L9 - 1 1 turns No. 22 enameled wire on a L1 1 - 8 turns No. 22 enameled wire on a 

T50-2 core. T50-6 core. 


Modulation Methods 205 




The transmit mixer used by KL7IAK. 


As modified with the emitter degen- 
eration, the amplifier appeared to be 
thermally stable. The amplifier chain 
was run at full output for a five-minute 
period. When the drive power was re- 
moved, the collector current in the 
output stage was SO mA, and quickly 
decreased to the previously established 
15 mA. A two-tone test on the total 
power chain produced IMD products 
over 30 dB below each output tone. The 
output power during the test was 
2.5-watts PEP. 

The output-amplifier chain was built 
in Oregon where instrumentation was 
available for careful evaluation. The 
experience with thermal runaway was 
very impressive for the writers, suggest- 
ing that emitter ballasting is a necessity 
for any Class AB ssb amplifier. This 
includes units for QRP operation as well 
as the higher power versions. 

If this transmitter is built for other 
bands, the circuits must be changed. 
The narrow -band power chain shown in 
Fig. 37 could be adapted to any of the 
amateur bands from 1 .8 to 30 MHz. 
However, a more modern approach 
would be to utilize broadband designs. 

Shown in Fig. 38 is the circuit of an 
amplifier suitable for following the 
mixer of Fig. 35. Network Z2 is a 
double tuned circuit for the band of 
interest. The component values for 
these filters are listed in the computer- 
generated tables in the appendix. The 
2N5179 amplifier is flat into the vhf 
spectrum with a gain of almost 20 dB. 

A broadband Class A power ampli- 
fier is presented in Fig. 39. This circuit 
has two stages, provides gains up to 36 
dB. and will deliver an output of 1-watt 
PEP or cw. The transistors in the output 
stage should be fastened carefully to a 
suitable heat sink, since the current in 
the final is 250 mA. 

The driver in the rf-power chain is 
identical to the amplifier described in 
Fig. 38, except that the collector cur- 
rent is higher. The output amplifier uses 
a pair of 2N3553s in parallel. Emitter 
degeneration is used in this amplifier for 
bandwidth extension. The emitter resis- 
tors further ensure that the dc current 
in the transistors is divided equally. Also 



Fig. 38 - Bandpass filter and broadband preamplifier for the rf-output chain of the universal ssb 
transmitter. This circuit follows the transmit mixer of Fig. 35. T2 consists of 1 0 bifilar turns of 
No. 28 enam. wire on an Amidon FT-23-43 toroid. Z2 is a bandpass filter from the tables in the 
appendix. 



Fig. 39 - Broadband Class A power amplifier for the universal ssb transmitter. The output 
power is 1 -watt cw or PEP linear. This circuit is suitable to follow the filter and amplifier of 
Fig. 38 for any band from 1 .8 to 30 MHz. Heat sinks should be used on all three transistors 
in the amplifier. The insert shows a modification which is suitable for 0.5 watt of output. 
This filter should be followed by a low-pass filter for the band of application. Suitable filters 
were described in chapter 4. T 1 is 1 2 bifilar turns No. 30 enameled wire on an Amidon 
FT-37-61 ferrite toroid. 


206 Chapter 8 






Breadboard version of a 1-watt output Class A broadband power amplifier. The circuit provides 
over 30 dB of gam over most of the hf region. Heat sinks are used on the parallel 2N3553 out- 
put amplifiers. 


+12V 



BAND 

Y1 

MHz 

iy 

LI 

L2 

Cl 

<pF) 

C2 

(pF) 


40M 

11 

16-16.3 

24 ts 

2 ts 

100 

75 


15M 

17.5 

12-12.5 

22 ts 

2 ts 

50 

47 


10M 

14 

19-19.5 

21 ts 

2 ts 

75 

47 

* = Nominal Cl required. 

6M 

36 

41-41.5 

13 ts 

2 ts 

30 

15 

Z 3 Filter from tables. 

160M 

5.8 

10.8-11 

45 ts 

4 ts 

100 

100 

^ " F out 


Fig 40-Circuit lor a heterodyne conversion system lor the VFO. A 5- to 5.5-MHz VFO such as that 
shown in Fig. 36 is heterodyned to the needed output frequency for operation on any amateur band 
trom 1.8 to 50 MHz Note that the values given in the table for Cl are nominal values A slightly larger 
mica compression trimmer should be used. All coils for the crystal oscillator are wound on Amidon 
T50-6 toroid forms Z3 is a 2- or 3-pole bandpass type trom the appendix (see text). The SN-76514 mix- 
er 1C has been reidentified as TL-442-CN by Texas Instruments, It may be procured under either part 

number. 


shown in Fig. 39 is an adaptation of the 
circuit using a single 2N3553. This 
circuit should provide identical gain and 
bandwidth, but will have an output 
power of only 1/2 watt. 

The broadband amplifier was evalu- 
ated for IMD while usinga pair of signal 
generators at 14 MHz, and a spectrum 
analyzer. The output intercept was 
+43.5 dBm. When the amplifier was run 
at 1-watt PEP output (+24 dBm per 
tone) the IMD was 39 dB down. The 
maximum gain was 36 dB in the 3.5- 
and 7-MHz bands. The gain was down 
to 34.5 dB at 14 MHz and was 29 dB at 
29 MHz. If the transmitter is built for 
the 6-meter band, it is suggested that 
the power amplifier used in the previ- 
ously described 144-MHz dsb trans- 
mitter be used. The output network 
must be altered. 

The broadband amplifier (Fig. 39) 
should be followed by a low-pass filter. 
Half-wave filters are suitable (see 
chapter 4). 

When the transmitter is used on 
bands other than 20 or 80 meters, a 
different VFO system is needed. A 
solution is to use a heterodyne VFO. 
Shown in Fig. 40 is a schematic for a 
proposed system that could be built for 
any of the bands from 1 .8 to 50 MHz. A 
5- to5.5-MHz VFO is used. Its output is 
heterodyned to the suitable injection 
frequency. An SN76514 double- 
balanced mixer 1C is used. A crystal- 
controlled oscillator is employed as the 
other input to die VFO mixer. Values 
are given for the oscillator components 
for all bands. 

The output of the premixer (Fig. 40) 
must be filtered well in order to sup- 
press spurious responses. A double- or 
triple-tuned circuit is used. The circuit 
should be terminated in 50 ohms at the 
output. 'Hie input termination should 
be 600 ohms to match the output of the 
SN765I4. Filters values are given in the 
appendix. They are designed for a 50- 
ohm termination at each end. The 
methods to adapt them to other termin- 
ations are also presented. Either 2- or 
3-pole filters may be used. For most 
cases the double-tuned circuit will be 
sufficient. The 3-pole filters are prefer- 
able for the 1 0- and 15-meter bands. 

The transmit mixer requires an injec- 
tion power of +4 dBm. If this level is 
not available at the output of the filter, 
it may be increased by means of a 
broadband amplifier. 

While the circuit shown in Fig. 40 
has not been built, we feel that it should 
present no problems. Two other proj- 
ects in the book use a similar circuit in a 
virtually identical application. No prob- 
lems were encountered with those 
designs as long as the proper filter 
terminations were used. 

A control system for the ssb exciter 
is shown in Fig. 4 1 (see chapter 7). All 


Modulation Methods 207 




switching functions are done with tran- 
sistors except for the antenna section 
which utilizes a relay. A delay is built 
into the system to ensure that the 
antenna relay is closed prior to genera- 
tion of rf from the transmitter. Shaped 
keying is also included. Two pnp transis- 
tors are used for switching. These sup- 
ply the +12T and +I2K (keyed) lines in 
the transmitter. These transistors should 
be capable of switching up to 1 ampere. 

At this writing only one of these 
transmitters has been built - the origi- 
nal KI.7IAK unit. It has been highly 
successful on both ssb and cw. 

It should be emphasized that the 
universal ssb system described in the 
preceding pages is an advanced project. 
Although well within the capabilities of 
the amateur with construction experi- 
ence, it should not be attempted by the 
beginner. There is no printed-circuit 
layout information available on any of 
the projects described in this chapter. 



The right-hand board contains the carrier oscillator and balanced modulator of the KL7IAK 
transmitter. The circuit at the left is the audio section. Two stages of audio were used, but 
later found to be unnecessary. 


Fig. 41 Control system for the universal ssb system. This circuit provides automatic T-R switching on cw and push-to-talk operation on ssb. 
The design details of these control systems were presented in Chapter 7. 



208 Chapter 8 



Chapter 9 


Field Operation, Portable Gear 
and Integrated Stations 


^/lost of the equipment described in 
this book is suitable for field use, be the 
application one of weekend camping, 
mountain climbing, Inking, boating, or 
long-term vacationing abroad or in the 
USA. The exact nature of the material 
taken afield will depend to a large 
extent upon the environment in which 
the gear shall be used. In more definitive 
language, the equipment must be de- 
signed for extreme compactness in some 
instances, and must be capable of opera- 
ting from batteries. The backpacker and 
hiker are especially mindful of the 
foregoing requirements, and would add 
to their list of accessories a lightweight 
antenna system, headphones, key, 
and/or microphone. 

The lakeshore or river-side camper 
might elect to carry larger, more power- 
ful radio equipment with him. lie could 
utilize the automobile battery or a 
gasoline-powered generator to obtain 
the needed source of energy. His an- 
tenna system could be more rugged and 



Low -power station equipment can be used in 
place of commercial gear when the QRP 
challenge inspires the operator. On the left 
side of the operating position is the W1CER 
40- and 20-meter 10-W station. The power 
supply and Transmatch for the homemade 
setup are on the shelf behind the QRP gear. 


elaborate than that used by the back- 
packer. 

Those who operate from motels or 
hotels, stateside or in some distant land, 
would be more apt to employ an ac- 
opcrated power supply which was com- 
patible with the line voltage and fre- 
quency in the area where operation was 
planned. However, a rechargeable 
battery might also be included in tire 
travel kit for use at times when local 
power failures occur - and they do in 
many foreign countries! 

There is a mystique connected with 
portable operation, for in many 
instances the amateur is using home- 
made equipment which was tailored to 
the application. Furthermore, low 
power is employed much of tire time, 
and conditions are seldom ideal with 
respect to operating conveniences. Being 
heard, and having other station opera- 
tors copy your signal solidly, not only is 
a measure of your station effectiveness, 
it’s a self-satisfying feather in the cap of 
the designer/ operator. “Doing it the 
hard way" does not necessarily denote a 
twinge of masochism. Rather, it proves 
that QRP gear is worth its weight when 
applied properly. 

Equipment Characteristics 

The environment at the site of porta- 
ble operations is of major importance to 
the designer. For example, the moun- 
tain climber will encounter extremes of 
cold, which can affect the performance 
of his equipment if certain design steps 
aren’t taken. His transceiver and related 
apparatus must be small and light of 
weight - and rugged if it is to suit his 
particular needs correctly (more on this 
subject later). 

The camper needs equipment that 
can function properly in damp weather. 
It should be reasonably immune to dirt 
and temperature extremes, and requires 


a quality of ruggedness which most 
home -station gear need not have. 

The foreign traveler will often 
choose compact equipment, owing to 
the inconvenience of lugging a large, 
heavy commercial transceiver. Light- 
weight, compact gear can be carried 
aboard an airplane without the penally 
of being “overweight." The latter can 
become rather expensive! Also, the 
station equipment is less likely to be 
damaged if kept out of the hands of 
baggage men during air travel. Being 
able to take the package of radio equip- 
ment to one’s seat on the plane will also 
prevent misr outing of the parcel to 
some destination other than the 
intended one! The writers recall an 
unhappy event that found the entire 
DXpedition radio package missent to 
Trinidad, when the operators and their 
personal effects were destined to land 
on Barbados (W1KLK, W1CKK and 
W1CER). Not only did the radio gear 
become lost temporarily, the suitcases 



Solid-state QRP station used by W1CER at 
ZF1ST. A backup keyer and the station 
power supply are at the left. The top-center 
unit is the 40- and 20-meter 10-W cw trans- 
mitter. A 160- through 15-meter superhet- 
erodyne receiver is below the transmitter. 

A small speaker flanks the receiver to the 
right, then comes a homemade keyer with a 
commercial paddle. 


Field Operation, Portable Gear and Integrated Stations 209 


1 





Portable operation can take place from a 
makeshift table. Here, the trunk of a VW 
fastback is used as an operating position. 

A 12-V battery is used to power this 1-W 
80- and 40-meter cw transceiver. A home- 
made keyer is visible at the right. This 
station was used during an ARRL CD Party 
by W1CER during a New Hampshire camping 
trip. 


containing clothing and cosmetics van- 
ished at the same time. The errant 
luggage turned up a few days later at the 
seaside resort on Barbados, but the 
radio equipment had been damaged 
severely. The lesson learned was that 
hand-carried QRP equipment was the 
better choice for traveling by air! 

Tent Camping 

Most "purist” campers who dwell in 
tents will not be situated where ac 
power is available. Chances are that they 
will not be close to an automobile, 
which will rule out “snitching” equip- 
ment power from a car battery. Not 
many ardent campers will justify pol- 
luting the serenity of the wilderness by 
using a noisy, gas-gulping power plant. 
Therefore, various types of battery 
power supplies become the order of the 
day. Some camper/amateurs use series- 
connected 6-volt lantern batteries to 
obtain 12 volts for the QRP gear. Others 
employ Gel-Cell or NiCad batteries. Still 
others obtain good results with flash- 
light cells connected in series to provide 
the required operating voltage. The 
choice is based usually on what’s avail- 
able at the time, and on the power 
consumption of the field equipment. 
Another excellent power source is a 
12-volt motorcycle battery, or two 6- 
volt ones hooked in series. If the auto- 
mobile is close enough to the campsite 
to permit occasional recharging of the 
batteries, NiCads, Gel-Cells and motor- 
cycle batteries are the best bet. If dry 
batteries are used exclusively, it’s wise 
to carry enough spares to bracket the 
arrival and departure dates adequately. 

Assuming that battery power is used, 
the equipment should not consume 
more than a few hundred milliamperes 
with everything running. The cw operat- 
ing mode will probably be the most 
efficient one. Effective communications 
should be possible from 160 through 10 
meters while using power levels from 
0.5 to 3 watts, assuming that a reason- 


ably good antenna is employed, and 
that band conditions are suitable. A 
good day- and night-time frequency 
combination is 20 and 40 meters. Propa- 
gation on those bands will permit 
round-the-clock operating, most of the 
time. 

Dipole antennas are among the easi- 
est to transport and erect when camp- 
ing. They can be supported by tall trees 
or cliffs - erected as inverted Vs, 
sloping dipoles, or in a traditional for- 
mat - horizontally. A bow and arrow is 
useful when erecting antennas, for it 
permits a pilot line to be fired and 
snaked through a treetop, preparatory 
to pulling the antenna aloft with a 
heavier line. Those skilled with a spin- 
ning rod can shoot a quarter-ounce prac- 
tice lure or sinker over a treetop, then 
pull the antenna line up by hooking the 
monofilament fishing line to the main 
one. 

An inexpensive but good antenna- 
support line is the Nylon type which 
can be bought in many hardware and 
discount stores in the USA. A 500-foot 
roll will last a long time and will cost 
less than $3. The writers prefer the 
small-diameter kind which has a tensile 
strength of 100 pounds or greater. When 
the campout is finished, the cord can be 
placed back on the spool for use an- 
other time. 

It may be necessary to use the radio 
equipment on the ground, as some 
campers do not carry tables and chairs 
afield. Therefore, the equipment should 
be sealed reasonably well against sand, 
moisture and insects. When not in use, 
tire gear should be wrapped in plastic 
food bag:, to keep it dry and clean. A 
shady operating position is best, as the 
operator will be more comfortable, and 
tire equipment will not be subjected to 



A sloping dipole strung near the seashore 
makes an effective antenna for QRP DX- 
peditions. Shown here is the ZF1ST/W1CER 
40-meter dipole used on Grand Cayman 
Island (Spanish Bay Reef). Power output 
from the transmitter was 7 watts, and RST 
589 reports were received from JA stations 
during the operation. 



One advantage of QRP gear is that it doesn't 
occupy a great deal of room. In this photo- 
graph the station (HW-7 and a homemade cw 
transceiver) occupy one corner of a tent dur- 
ing a camping trip. The equipment is powered 
from a 12-V battery. 


extremes of heat from the sun. The 
latter can cause expansion of critical 
tuning mechanisms (trimmers and coil 
slugs), leading to degraded performance 
and a need to readjust the circuits. 

Since accessory equipment for camp- 
ing and out-of-country operations is 
similar, that subject will be covered 
singly, later in this section. Generally 
speaking, the same kinds of antennas are 
adequate for both applications. 

QRP DXpeditioning 

There is probably no greater thrill in 
amateur radio than that of being DX 
with QRP equipment. W1CER has made 
several trips to islands in the West Indies 
for the purpose. Much of the work was 
done as 8P6EU from Barbados, with 
XYL Jean. W1CKK/8P6FJ, as a second 
operator. Other operations took place 
from Grand Cayman Island as ZF1ST. 
Propagation from that part of the world 
is superb to the USA and Europe, 
making it practical to employ low- 
power transmitting gear. The antennas 
have always been half-wave dipoles 
(coax fed) which were erected as 
"slopers” at whatever height was possi- 
ble. Because salt water constitutes a 
superb ground medium, the antennas 
were slug over the seashore to assure 
best performance. The maximum trans- 
mitter output power used was 7 watts. 
Much of the work was done, however, 
with 1-1/2 to 2.5-watts output. The 
primary bands of operation have been 
40, 20 and 15 meters. Cw was the 
operating mode. 

Solid QSOs were had with many 
amateurs from Europe, South America 
and the USA. From Grand Cayman 
during October of 1974, a number of 
Japanese stations were worked on 40 
meters at sunrise, local time (1000 
GMT). Power output was 7 watts, and 
the antenna was a sloping dipole, the 
center of which was 15 feet above 
ground! Signal reports both ways were 
RST 589. ZLs and VKs have been 


210 Chapter 9 



worked with 2 watts and a sloping 
dipole from Barbados. The period was 
early sunrise, and the band was 20 
meters. Contacts like that are the excep- 
tion rather than the rule, but they can 
be made with QRP equipment. Some 
signal enhancement from 8P6EU to 
Oceania probably resulted from having 
the 20-meter sloping dipole facing west 
on the western side of the island. 
Furthermore, a 30-foot coral cliff was 
behind the antenna (cast), helping to 
effect some directivity. 

There are many fine Caribbean 
islands from which to operate. Prior 
familiarity with government regulations 
is recommended, lest an amateur arrive 
and not be granted operating privileges. 
On Barbados a license can be acquired 
only in person. One must present his 
U.S. license to the Government Electri- 
cal Inspector, Old Hospital Bldg., 
Bridgetown. The fee for 12 months is 
nominal, and the license can be renewed 
yearly by mail. A Caymanian reciprocal 
permit can be obtained by mail if the fee 
is sent alongwith a photocopy of the U.S. 
license. The call will be your U.S. one, 
slant ZF1. Applications must be ad- 
dressed to Her Majesty’s Postmistress, 
Licensing Division, Post Office, George- 
town, GCI, BWI. The fee for one year 
on Grand Cayman is fairly stiff, and the 
rate changes from time to time. 

It is wise to write to the local radio 
club on the island one plans to visit. 
Data can thus be obtained on Customs 
regulations and licensing. Some coun- 
tries will not grant a license, and others 
make it practically impossible to bring 
equipment in. On some of the islands 
one must post a bond which represents 
80 percent of the face value of the radio 
gear! Some amateurs have reported 
great difficulty getting all of the bond 
money returned at the end of their 
vacations! On some islands it takes a 
year or more to get a license, owing to 
government red tape. Be sure to check 
first; then make vacation plans. 

Accessory Equipment 

Campers and DXpedition types 
should anticipate equipment failures 
and prepare accordingly. If a backup 
station is not carried afield, spare parts 
and tools are a must. It is wise to 
conclude that most of these tilings will 
not be available once the operator 
reaches his destination. Radio stores just 
don’t exist in the back woods or on 
many West Indies islands, so take what 
you need with you. 

The following list of tools is sug- 
gested when space permits taking them 
along: Diagonal cutters, jack knife, elec- 
trical tape, screwdrivers, pliers (needle- 
nose and regular), small VOM, solder, 
soldering iron (battery operated), clip 
leads (6), cube taps, extension cord and 
hookup wire. These items will require 



A typical collection of equipment, antennas, 
spare parts and tools for a QrfP DXpedition. 
The materials in the picture were packed into 
the portable typewriter case seen at the rear 
of this illustration, then transported in hand 
to Barbados during operation as 8P6EU/ 
8P6FJ in 1973. 


very little additional space in the travel 
case, and may prove useful when setting 
up the station. Schematic diagrams of 
the equipment should also be taken 
afield, should troubleshooting be re- 
quired. 

Spare parts are important when 
operating portable, and a few compon- 
ents thrown into the tool kit could be 
helpful. Critical components, such as 
the PA and driver transistors of the 
transmitter, should be taken along as 
spares. Fuses, spare batteries, and a 
collection of capacitors are often handy 
when a failure occurs. The W1CER parts 
kit contains .001, .01, 0.1, 2, 10 and 
50-pF capacitors. Included also are rec- 
tifier diodes, high-speed switching 
diodes, general-purpose FET and bipolar 
small-signal transistors. Depending on 
the kind of circuit used, certain ICs are 
also included in the kit. 

Salt water, and the air near salt 
water, has a notable effect upon some 
kinds of amateur equipment. The keyer 
paddle will develop poor electrical con- 
tacts after a period of time near the sea. 
A machinist’s point burnishing file is 
handy for restoring the contacts of a 
key. Antennas mounted near the sea- 
shore for long periods should be coated 
with silicone grease to prevent corro- 
sion. This is especially true if aluminum 
tubing is used in the antenna system. All 
joints in wire antennas should be sol- 
dered rather than twisted together. That 
will prevent salt air and spray from 
causing poor connections. 

An SWR indicator is useful when 
afield: Some antenna pruning is usually 
required to provide a low SWR. A 
Transmatch can be taken afield for use 
with antennas that must accommodate 
more than one band of operation 
end-fed wires or a 40-meter dipole that 
will be used also on 1 5 meters. 

It is wise to check before traveling to 


a foreign land, to learn what the local 
power service is. Some parts of the 
world still use 25-11/ lines, while others 
use 50- or 60-llz lines at some unusual 
voltage amount. It may be necessary to 
carry a power converter when ac opera- 
tion is contemplated. Furthermore, the 
wall outlets in some countries are pretty 
strange to U.S. amateurs — an adapter 
may be necessary. 

One final word of advice: When 
abroad it is important to exhibit proper 
ladio conduct. Be especially courteous 
to the local amateurs you meet and talk 
with on the air. If you're operating from 
a hotel, use headphones rather than 
causing disturbances by operating with a 
loudspeaker. Be on the watch for inter- 
ference to TV sets and radios.- If the 
fault can’t be corrected, cease operating. 
Also, a secondary frequency standard 
should be included with your radio gear. 
Straying out of an authorized amateur 
band could be embarrassing and expen- 
sive. Some foreign governments require 
that you have a crystal-controlled secon- 
dary standard before they will allow 
you to operate. A 100-kHz calibrator is 
usually adequate. 

Wilderness Operation 

The preceding section dealt with the 
problems encountered during operation 
at camping and DX locations. While 
such activities are certainly glamorous, 
especially for the DX operator, other 
portable ventures can produce similar 
rewards. 

For over a decade a dominant activ- 
ity at W7ZOI has been operation in 
connection with mountaineering and 
backpacking trips. Tire equipment re- 
quirements are different than they 
would be for other portable stints. All 
of tire equipment must be carried on the 
back of the operator. This presents no 
problem if the walk is short and the 
purpose of the trip is specifically for 
“hamming" for a short duration. How- 
ever. when the operation is secondary to 
a physically more ambitious goal, such 
as reaching the summit of a major peak, 
the criteria change. 

There is a philosophy practiced by 
backpackers when assembling equip- 
ment for an extended outing. Simply 
stated, it is “Worry about the ounces - 
the pounds will take care of them- 
selves.” This approach must be ex- 
tended to the design of any radio 
equipment that will be taken aloft. 

The equipment should be designed, 
built and tested in the winter months. 
The gear is then ready when spring 
arrives. Hasty construction just before a 
trip invites equipment malfunction. 

The primary consideration is weight 
rather than volume. Excessively dense 
packaging may be entertaining for the 
builder. However, if it makes the equip- 
ment less reliable and versatile due to 


Field Operation, Portable Gear and Integrated Stations 21 1 



Author Hayward is seen here during a moun- 
tain trip on which he took the 40-meter Ultra 
Portable Transceiver described in this chapter. 
The battery pack is in his jacket pocket. 

component crowding, it should be 
avoided. A reasonable size for a com- 
plete rucksack station is 2 X 5 X 7 
inches. This allows ample room for 
circuitry while keeping construction 
straigh tforward . Batteries should be 
external. 

The heaviest items to be considered 
are the batteries and antenna. The size 
of the batteries required will depend 
upon the power level of the transmitter 
and on the expected period of opera- 
tion. This brings us to a major con- 
straint keep the power as low as 
possible. It is difficult to say how low it 
should be. At W7ZOI. the portable 
power levels used, mainly at 7 MHz 
(cw), have ranged from 8 W down to 
250 mW of output. Our impression is 
that an output of 0.5 to 1 watt is near 
optimum for use in die contiguous 
states and in the less remote parts of 
Canada. This allows the use of Penlight 
cells or NiCads for short operating 
periods. Higher powers are useful in the 
more remote areas or for contest work. 

Temperature extremes can have a 
dramatic effect on equipment perfor- 
mance. Oscillator instability is one 
common problem. In one cold experi- 
ence (Mt. Adams in Washington State), 
a germanium transistor oscillator would 
not start. During another trip with vhf 
gear, low temperatures caused severe 
detuning of a frequency-multiplier 
chain. 

There are a number of factors to 
consider when designing for tempera- 
ture extremes. Semiconductors with 
wide operating-temperature ranges are 
suggested. Crystal control is recom- 
mended. While this is not mandatory for 

212 Chapter 9 


a receiver, it is highly desired for a 
transmitter. Frequency accuracy is of 
importance if schedules are made with 
odier stations. The crystals should be 
built into the equipment and selected 
with a switch. Loose crystals are easily 
lost. Critical tuned circuits should be 
avoided. Finally, the equipment should 
be tested under severe conditions. While 
the home freezer can be used, it’s 
usually more fun to take the equipment 
into tlie field when the first winter 
storm arrives. 

There are other criteria that might 
be applied when designing equipment 
for mountaineering application. They 
might seem extreme, but have been 
found useful. First, it is desirable that 
the equipment be operable with gloves 
being worn by the operator. Sharp edges 
should be removed. This lessens the 
possibility of tearing holes in a tent or 
sleeping bag. It's worthwhile to build 
die equipment so diat it may be opera- 
ted in the dark. This is particularly 
useful during winter trips when the rig 
must reside inside the sleeping bag with 
die operator. Bodi must be warm to 
function well. Most batteries will de- 
crease in output voltage and energy 
capability when cold. Provision should 
also be made to keep the battery pack 
warm. 

As an aside, a winter trip on skis or 
snowshoes is an especially enjoyable 
time for taking the radio gear along. 
Owing to the long nights, it is often 
necessary to spend from 1 2 to 1 4 hours 
at a stretch in a mountain tent or snow 
cave. The ham gear helps to pass die 
dme. Also, the possibility of being 
stranded by a change in the weather is 
greater. Reliable communications capa- 
bility could be very valuable. 

Questions asked by the prospective 
portable operator are, “What band and 
mode to use?” First, cw is preferred 
over ssb. The equipment tends to be 
more reliable owing to the simplicity. 
The narrower information bandwidths 
help. However, the operator should be 
proficient with conversational cw - diat 
is, he should be able to copy the code 
without having to put anything on 
paper odier than logging information 
and a few notes. Physical strain and the 
effects of a harsh environment make 
normal operation difficult. The less pro- 
ficient cw operator should consider ssb 
or dsb equipment. 

The choice of frequency is difficult, 
and partially subjective. For summer 
operation, 40 meters is ideal. The band 
remains open for short hops during the 
daylight hours and well after sundown. 
Eighty meters is better for winter use. 
Noise levels are too high for 80-meter 
effectiveness in the late summer. The 
20-meter band is excellent for the oper- 
ator with an interest in evaluating un- 
usual locations for DX effectiveness. In 


Field Day contests where the writers 
have participated with QRP, 20 and 40 
meters have produced the largest num- 
ber of contacts. 

The vhf bands are, in many ways, 
the optimum bands to explore. It is at 
diese frequencies where the real benefits 
of a high, mountain-top location are 
obtained. Equipment is more complex, 
but not unreasonable. Sideband or cw 
are still the recommended modes. Cw 
has the edge for long-haul work. How- 
ever, more vhf operators will be comfor- 
table with phone. 

Antennas can be a problem in the 
mountains. The standard carried in the 
W7ZOI rucksack is a 7 -MHz dipole with 
a 40-foot transmission line of RG-174. 
While a larger size cable would have less 
loss, the added weight would be intoler- 
able. The 40-foot cable mentioned has 
0.7 dB loss at 7 MHz. The center 
insulator for the dipole is made from a 
scrap of pc board (unclad). The antenna 
is always operated in the inverted-V 
configuration. This has the advantage 
that only one support is needed. Trees 
are ideal when available. Above timber- 
line, a small telescoping whip antenna is 
carried. The unit used collapses to 14 
inches and is 12 feet long when ex- 
tended. It has always been more effec- 
tive to use the whip as a center support 
for the dipole than to load it against a 
grpup of radials. This whip was a surplus 
item and the source is unknown. A good 
substitute would be a long fishing rod. 
Fiberglass rods up to 20 feet long are 
available through the Sears f ishing and 
Boating Catalog. 

The usual methods for putting a line 
into a tree are effective. A lightweight 
fishing line is preferred over a heavier 
nylon cord. Also, it is wise to carry a 
1 -ounce weight with the other gear. On 
one occasion, one of the writers found 
himself without an adequate weight to 
tie to the end of a nylon cord. Not 
wanting to miss out on Field Day, a 



A close-up view of W7ZOI‘s hands holding 
the 40-meter transceiver. Despite the gloves 
he was able to operate the built-in key 

lever during cw transmissions. 



Transmatch was attached to the rope 
and hurled aloft. 

A small beam is recommended at 
vhf. This may be lashed to an ice ax for 
above-timberline operation. The an- 
tenna should be capable of easy 
assembly with a minimum of loose parts 
needed. 

No matter what equipment is used, 
or what the goals of the operator are, 
respect should be maintained for other 
mountain travelers. Rambling into coun- 
try which is devoid of roads or even 
trails offers an escape from the daily 
pressures and routine that have become 
a dominant part of our society. Back- 
packing and mountaineering are increas- 
ing in popularity and. unfortunately, 
the “wilderness” is often an area with a 
number of visitors. The last thing a 
fellow hiker wishes to hear is the blare 
of cw rushing from an overdriven 
speaker. Headphones should always be 
used! 

Finally, the radio amateur who 
carries his hobby into the back country 
should be prepared for an occasional 
emotional dilemma. Should he com- 
promise his hiking or climbing goal in 
order to get on the air or should he 
pursue the primary goal? In this age 
dominated by high technology, the 
answer is obvious. Climb the mountain! 

QRP Operation 

Although portable operation has 
been the motivation for the work of the 
writers respective to QRP equipment, 
this is not typical. The more common 
QRP operation occurs from the home 


station. The operator's motivation is to 
add excitement and adventure to con- 
tacts that would otherwise offer minor 
challenge. Much of the present QRP 
popularity results from ready availabil- 
ity of commercial equipment at reason- 
able prices. Fortunately, the excitement 
has spurred many amateurs to build 
their own gear, allowing them to gain 
doubly from their operating activities. 

The criteria for success with QRP 
gear are not all that different titan they 
are for high power. The key is in the 
antenna system and in a wise choice of 
operating frequencies for a given time of 
the day or year. These decisions are 
more critical with low power. 

There are a few operating techniques 
that can aid the QRP operator. Gener- 
ally, he will be more successful if he 
calls other stations rather than calling 
CQ. Often, it is better to call a station as 
he is finishing a contact rather than 
answering a CQ from a loud station. 
Another trick is to add some additional 
information to a call, letting the fellow 
on the other end know that there is a 
reason for the signal being weak. This 
can be successful even when calling CQ. 
However, it is usually not enough to 
tack a "QRP" on to a CQ. To some 
operators, this merely implies that the 
station signing “QRP” is running less 
that 100 watts, only 10 dB down from 
the legal limit. A much more effective 
format is CQ CQ de QRP 1 watt, 
W7ZOI W7ZOI and so forth. The 
writers feel that these methods should 
not be applied except for output powers 
of less than 1 or 2 watts. 


An excellent time for the QRP oper- 
ator to make a large number of contacts 
is during contests. Here there are a 
larger number of stations available to be 
worked. Of greater significance, they are 
anxious to work anyone they can, and 
will not be upset with a less than 
ear-shattering signal. 

One of the best contests for QRP 
work is Field Day, for a large number of 
similar stations are active during the 
same period. This has been aided by the 
individual listing of low-power stations 
in the QST results. A club Field Day 
using QRP is an interesting and unusual 
experience. TERAC (Tektronix 
Employees’ RAC, K7AUO) has partici- 
pated in the QRP category for several 
years. While many of the more competi- 
tive, contest-oriented members have 
avoided the activity, others with general 
interests have participated and have 
enjoyed low-power work. A QRP Field 
Day tends to be a more relaxed affair. 
This is aided by the conspicuous 
absence of the roar of a generator. 

Although QRP operation may seem 
casual, there are some who have become 
accomplished in this area. Many opera- 
tors have achieved WAS and WAC with 
quite low powers, and a few have 
qualified for DXCC with less than 5 
watts of rf output. Generally, tire more 
successful QRPers are cw enthusiasts. 

An interesting experiment is to 
attempt contacts with as little power as 
possible. Minimum-power experiments 
were performed during a number of 
contacts between W7ZOI and WA6YVT 
in 1969 and 1970 on 40 meters. To 
attach some legitimacy to the contacts, 
a strict format was established. Contact 
was established initially with an output 
of I to 3 watts from W7ZOI. If the 
reports from WA6YVT (in the Los 
Angeles area) were favorable, the output 
power would be decreased. A step atten- 
uator was used in a matched 50-ohm 
antenna system to ensure that the out- 
put power at W7ZOI was well defined. 
At each power level, an arbitrarily 
chosen 4- or 5-letter word would be 
sent. The word was repeated several 
times. WA6YVT would then repeat the 
word to confirm that information had 
actually been exchanged. It was not 
possible for a vivid imagination to serve 
as a substitute for actual copy. 

While experiments were conducted 
to evaluate the power levels that would 
be suitable for portable equipment, they 
turned out to be generally interesting. 
In nearly all cases where the attenuator 
was put into the transmission line, 
information was exchanged at 100-mW 
output. Often 50 mW was successful. 
The lowest power producing a real 
exchange of information was 2.5-mW 
output. Immediately after that contact, 
the output power was confirmed with a 
high-frequency oscilloscope. One con- 



Photograph of the W7ZOI home station. All of the amateur equipment and test gear is home- 
made. The operation position serves double duty by also being a workbench. 


Field Operation, Portable Gear and Integrated Stations 213 



PRODUCT 

DETECTOR 


1QO>uF 
19V , 


02 

2N3904 


MPF102 


I-H2V |_ 

-F AMP 



1 +12V 


>4700 \ 

>27. 9 


4 

7 .005 

r -)h 

>120 

_W , 

4 r- 

I? — 1L T 



U1 

MC1350P 8 

■ i. - T,-n 

— |L6 4.4 MH. 

3IL 


AF AMP. 


$ M - SILVER MICA 


EXCEPT AS INDICATED, OECIMAL VALUES OF 
CAPACITANCE ARE IN MICROFARADS I »F I 
OTHERS ARE IN PICOFARADS I eF OR J.J.F)', 
RESISTANCES ARE IN OHMS ; 

► ■1000. M.IOOO 000. 


400 mw BUFFER 


? $ M>, 


Fig. 1 - Schematic diagram of the VFO and receiver portions of the 7-MHz transceiver. Fixed-value capacitors are disk ceramic unless otherwise 
noted. Fixed-value resistors are 1/4- or 1/2-W composition. Variable capacitors without part numbers can be mica compression trimmers 
(surplus Teflon or ceramic trimmers were used in the authors' unit). 

Cl — 80-pF air variable. enam. wire. L9 — 43 turns No. 26 enam. wire on 

FL1 — Crystal filter, ladder type (see text). L5 - 34 turns No. 26 enam. wire on T50-2 toroid core. 

LI - 30 turns No. 26«nam. wire on T50-2 toroid core. LI 0 - 29 turns No. 24 enam. wire on 

T50-2 toroid core. L6 — 36 turns No. 26 enam. wire on T50-2 toroid core, tapped 7 turns 

L2 — 36 turns No. 26 enam. wire on T50-2 toroid core. from ground end. Coat with Q dope. 

T50-2 toroid core. L7 — 5-turn link over L6, No. 26 enam. T1 — 12 trifilar turns No. 30 enam wire on 

L3 — 3-turn link over L2 winding, No. 26 wire. Amidon FT37-61 ferrite toroid core. 

enam. wire. L8 — 8-turn link over L9. No. 26 enam. U1 — Motorola 1C. 

L4 - 8-turn link over L5 winding, No. 26 wire. VR1 - 6.2 -V, 400-mW Zener diode. 


L9 - 43 turns No. 26 enam. wire on 
T50-2 toroid core. 

LI 0 - 29 turns No. 24 enam. wire on 
T50-2 toroid core, tapped 7 turns 
from ground end. Coat with Q dope. 
T1 - 12 trifilar turns No. 30 enam wire on 
Amidon FT37-61 ferrite toroid core. 

U1 — Motorola 1C. 

VR1 - 6.2 -V, 400-mW Zener diode. 


214 Chapter 9 









Front of the 7-MHz superheterodyne QRP 
transceiver. The jack at the left accommo- 
dates an electronic keyer. Audio output is 
taken from the side panel (right). The case 
measures 3X5X7 inches. 

elusion was that the method was an 
accurate means for evaluating the over- 
all condition of the path within an 
accuracy of 3 dB, far more accurate 
than an S-meter reading. However, the 
lowest powers were successful only 
when the propagation conditions were 
favorable and while noise levels were 
low. Similar methods would be useful 
for the study of vhf propagation. Addi- 
tionally, the weak-signal experience 
would be valuable to the operator with 
an interest in modes such as moon- 
bounce. 

A Superheterodyne CW Transceiver 
for 7 MHz 

For the beginning experimenter with 
an interest in QRP and portable opera- 
tion, a direct-conversion transceiver is 
ideal. Construction is straightforward, 
owing to the simplicity of design. When 
a higher level of performance is desired, 
especially in the receiver, it is better to 
build a superheterodyne system. 
Transceive operation is still desirable for 
some applications. Contests such as the 
ARRL Field Day are an example. 

The transceiver described in this 
section is based upon the preceding 
design criteria. The unit tunes a 
100-kHz segment of the 40-meter cw 
band. A full transceive type of trans- 
mitter with an output of 1 .5 watts is 
employed. The receiver selectivity is 
provided by a homemade 3-pole crystal 
filter of the lower side-band ladder type. 
The bandwidth is 250 Hz and the 
rejection of the undesired sideband is 
approximately 60 dB. A completely 
electronic T-R system is included, pro- 
viding smooth, transient-free control. 
Owing to the subtleties of the design, 
especially in the construction and align- 
ment of the crystal filter, this project is 
not recommended for the inexperienced 
experimenter. No pc information is 
available. 

Shown in Fig. 1A is the receiver 
section of this transceiver. The front 
end employs a dual-gate MOSFET as the 


mixer. A single tuned circuit (LI ) serves 
as the preselector. Four 1N914 diodes 
are used to protect the input. This 
would not be needed if the electronic 
T-R system were not employed. The 
mixer was not damaged when the diodes 
were omitted. However, they were 
included as a precaution against an 
improper termination at the antenna 
terminal. This could lead to high rf 
voltages at gate 1 of Ql. The output 
network of the mixer (L2-L3) is de- 
signed to present a termination of 125 
dims to the crystal filter. The design of 
the 4.4-MHz filter will be presented 
later. 

At the output of the crystal filter is 
a network (L4-L5) that presents a 50- 
dun termination to the filter. This is 
followed by an i-f amplifier which uses 
an MC1350P IC. This circuit provides a 
gain of approximately 40 dB, and allows 
for a gain variation of 60 dB. No age 
system is included in this transceiver. 
Receiver muting is realized by applica- 
tion of +12 volts to tlie arm of the 
manual gain control potentiometer. 

The i-f output is matched to 50 
dims (L6-L7) and then routed to a 
product detector utilizing four diodes. 
Originally, only two diodes were used. 
However, it was found that the im- 
proved balance obtained with four 
diodes provided less noise modulation 
of the BFO signal that found its way 
into the i-f amplifier. Note that the 
primary of Ti is balanced, being 
grounded only at the output of the i-f 
(L7). This also improved the balance. 
Such precautions would not be neces- 
sary if a less dense packaging format 
were used. Ground loops in a single 


continuous ground foil can cause these 
problems. They are avoided in systems 
employing a number of smaller circuit 
boards. 

BFO injection is provided by Q2. 
This oscillator is standard except that 
some means must be provided for 
adjusting the crystal to the proper fre- 
quency. All of the crystals used in the 
transceiver, including those in the filter, 
were cut for the same frequency. Exper- 
imentation may be required on the part 
of the builder to establish the proper 
capacitance across the crystal. 

A two-stage audio amplifier is used. 
Emitter degeneration is employed in 
both stages (Q3 and Q4) to ensure that 
linearity is preserved under large-signal 
conditions. The sidetone signal is in- 
jected into the base of Q3 during 
transmit periods. 

The VFO for the transceiver is 
shown in Fig. 1 B. A JFET is employed 
as a Hartley oscillator. Because the 
frequency is low (2.6 MHz), stability is 
excellent. The oscillator is tuned by 
means of an 80-pF air variable capaci- 
tor, which is driven by a Jackson 
Brothers vernier-drive mechanism. The 
fixed-value capacitance across the oscil- 
lator coil (L10) was chosen to provide 
the desired tuning range. A trimmer 
might be a useful addition to ease 
alignment. The oscillator is buffered 
with a two-stage amplifier, Q6 and Q7. 

The VFO is built in a small alumi- 
num box. This box is fastened securely 
to the front panel by means of standoff 
posts. Because all of the oscillator com- 
ponents are mounted securely to the 
smaller housing, mechanical stability is 
good. It was found that the transceiver 



Interior of the 7-MHz superheterodyne transceiver. At the center is the 2.6-MHz VFO compart- 

merit. The bottom pc hoard contains the receiver. At the top is the transmitter module. The 

small assemblies (2) at the right are the TR and sidetone boards. 


Field Operation, Portable Gear and Integrated Stations 215 



SWITCH 

012 

2N5906 


TO C4RRII 


Fig. 2 — Schematic diagram of the transmitter, T-R switching and sidetone circuits of the 7-MHz transceiver, Fixed-value capacitors are disk ceramic unless other- 
wise noted. Fixed-value resistors are 1/4- or 1/2-W composition Variable capacitors can be mica compression trimmers. The SN-76514 mixer 1C has been reidenti- 


fied as TL-442-CN by Texas Instruments. It may be 
L1 1 . L12— 23 turns No. 24 enam. wire on 
T50-2 toroid core. 

L13 — 20 turns No. 22 enam. wire on 
T50-2 toroid core. 

could be dropped 2 or 3 inches onto the 
operating table with no detectable fre- 
quency shift. 

Shown in Fig. 2 is the transmitter 
portion of the transceiver. Three circuit 
boards are employed. One board is used 
for the sidetone oscillator, with a 
second containing the electronic T-R 
switch. The third board contains the 
rest of the transmitter. 

The carrier oscillator (Q8) is a 
bipolar transistor operating in the Col- 
pitts configuration. To adjust the fre- 
quency to the center of the i-f passband 
it was necessary to add inductance 
(L19) and capacitance to the circuit. 
The crystal oscillator delivers 0.5 volt 
rms to the transmit mixer, U2. An 

216 Chapter 9 


procured under either part number. 

L14— 2-turn link over L13, No. 22 enam. wire. 
L15, L16 — 14 turns No. 22 enam. wire on 
T50-2 toroid core. 

LI 7— 42 turns No. 26 enam. wire on 

SN76514 was selected for the mixer, 
owing to the internally contained bias- 
ing resistors. An MC1496G could be 
used in this application. 

The output of the mixer is applied 
to a two-pole bandpass filter. The 
coupling capacitors between the resona- 
tors and into and out of the filter are 
critical and should not be substituted 
casually. The output of the filter is 
terminated in the 50-ohm input imped- 
ance of a feedback amplifier, Q9. The 
use of feedback is very useful where a 
well defined input immittance is 
desired. The buffer is followed by a 
driver, Q10. Both Q9 and Q10 are 
keyed by a pnp switch, Q12. 

The final amplifier, Q1 1 , uses a 


T50-2 toroid core. 

L18 — 4-turn link over 117, No. 26 enam. wire. 
L19— 40 turns No. 26 enam. wire on T50-2 
toroid core. 

Fairchild 2N4895 TO-5 type of transis- 
tor. A 2N5321 would function as well 
in this circuit. The output network is a 
half-wave filter. The output stage should 
have a small heat sink. 

The electronic T-R switch uses a pair 
of silicon switching diodes. The antenna 
is permanently connected to the trans- 
mitter. The receiver is also connected 
when the switching diodes are biased to 
an on condition. When the key is 
depressed, the 555 timer IC (U3) is 
triggered on. The output at pin 3 is then 
in a high state and supplies power to the 
transmitter carrier oscillator, Q8. The 
receiver is also muted, and the T-R 
diodes are reverse biased slightly. When 
the key is opened, U3 begins to time 


BUFFER 


X 

•TOO 01 

H 

S 2N2 


7-MH2 

BANDPASS FILTER 
aw-200«Hi 


TRANSMITTER BOARD 


SIDETONE 

BOARD 


T/R BOARD 


-o TO RCVR 


S M -SILVER MICA 











Close-up view of the receiver pc board. The front end is at the left, followed (right) by the 
three-pole ladder filter. The i-f amplifier is at the center of the board, with the product 
detector and audio amplifier at the upper right. In the lower right corner is the BFO. 


out. There is a short period before the 
receiver again becomes operational. 

Crystal-Filter Construction 

The crystal filter is shown in Fig. 
3A. All three crystals were at the same 
frequency with a maximum deviation of 
1 0 Hz. The crystals were measured using 
the methods outlined in chapter 5. This 
information was then used to design the 
Biter using die methods outlined by 
Zverev (see bibliography). The design 
predicted die values of the coupling 
capacitors and the resistances needed to 
properly terminate each end of the 
filter. The measured filter response was 


in excellent agreement with the design 
goals and no empirical changes were 
required for any of the values. The 
crystals used were surplus European TV 
color-burst types with a frequency of 
4433 kHz. 

Shown in Fig. 3B is the circuit of a 
two-pole filter of similar characteristics. 
This filter should be much easier to 
build on an empirical basis. Four 
crystals should be ordered at one time. 
An oscillator is built next. A frequency 
counter can be used to select die two 
crystals that are closest in frequency 
(Y4 and Y5). The other two are set 
aside for use in the BFO and carrier 


oscillator. Y4 and Y5 are soldered into 
place between the mixer and the i-f 
amplifier (Fig. 1A). Various values of 
02 are tried until the desired results are 
obtained. For a cw filter, a good starting 
point for C2 would be 470 pF. It may 
be necessary to change the terminating 
impedances. This can be done by experi- 
menting with die number of turns on 
L3 and L4. Although this procedure 
may sound a bit terrifying to the 
beginner, it is not difficult to obtain 
suitable results. Experimentation will be 
required though! 

A receiver was described in chapter 5 
which uses a filter with a single crystal. 
This circuit could be used in this trans- 
ceiver. However, the performance differ- 
ence between a single-pole response and 
diat realized widi two or three crystals 
is profound. 

The performance of this transceiver 
has been excellent. It has been used for 
portable and home-station QRP opera- 
don. Especially enjoyable has been the 
crisp response of the receiver and the 
smoothness of the control circuitry. 

Transceivers and Integrated Stations — 
Construction and Operation 

In diis section we are presenting 
construction projects for complete 
stations. Most are suitable for home- 
station use and operation from a por- 
table location.